Lipid Phosphate Phosphatase-2 Activity Regulates S-phase Entry of the Cell Cycle in Rat2 Fibroblasts*

Lipid phosphates are potent mediators of cell signaling and control processes including development, cell migration and division, blood vessel formation, wound repair, and tumor progression. Lipid phosphate phosphatases (LPPs) regulate the dephosphorylation of lipid phosphates, thus modulating their signals and producing new bioactive compounds both at the cell surface and in intracellular compartments. Knock-down of endogenous LPP2 in fibroblasts delayed cyclin A accumulation and entry into S-phase of the cell cycle. Conversely, overexpression of LPP2, but not a catalytically inactive mutant, caused premature S-phase entry, accompanied by premature cyclin A accumulation. At high passage, many LPP2 overexpressing cells arrested in G2/M and the rate of proliferation declined severely. This was accompanied by changes in proteins and lipids characteristic of senescence. Additionally, arrested LPP2 cells contained decreased lysophosphatidate concentrations and increased ceramide. These effects of LPP2 activity were not reproduced by overexpression or knock-down of LPP1 or LPP3. This work identifies a novel and specific role for LPP2 activity and bioactive lipids in regulating cell cycle progression.

Lipid phosphates are potent mediators of cell signaling and control processes including development, cell migration and division, blood vessel formation, wound repair, and tumor progression. Lipid phosphate phosphatases (LPPs) regulate the dephosphorylation of lipid phosphates, thus modulating their signals and producing new bioactive compounds both at the cell surface and in intracellular compartments. Knock-down of endogenous LPP2 in fibroblasts delayed cyclin A accumulation and entry into S-phase of the cell cycle. Conversely, overexpression of LPP2, but not a catalytically inactive mutant, caused premature S-phase entry, accompanied by premature cyclin A accumulation. At high passage, many LPP2 overexpressing cells arrested in G 2 /M and the rate of proliferation declined severely. This was accompanied by changes in proteins and lipids characteristic of senescence. Additionally, arrested LPP2 cells contained decreased lysophosphatidate concentrations and increased ceramide. These effects of LPP2 activity were not reproduced by overexpression or knock-down of LPP1 or LPP3. This work identifies a novel and specific role for LPP2 activity and bioactive lipids in regulating cell cycle progression.
The lipid phosphates, lysophosphatidate (LPA) 4 and sphingosine 1-phosphate (S1P) are present in biological fluids and activate cells through families of four G-protein-coupled receptors for LPA and five receptors for S1P (1). These receptors are coupled through G␣ i that decreases cAMP concentrations; G 12/13 that stimulates phospholipase D and Rho leading to stress fiber formation; and G q that activates phospholipase C, Ca 2ϩ transients, and protein kinase C isoforms (1). LPA and S1P receptors also transactivate epidermal growth factor and platelet-derived growth factor receptors (2,3).
Intracellular lipid phosphates also act as signaling molecules. For example, PA stimulates NADPH oxidase, protein kinase C-, phosphatidylinositol 4-kinase, phospholipase C-␥, and sphingosine kinase-1, increases Ras-GTP and inhibits protein phosphatase-1 (4 -6). PA can increase proliferation through the mammalian target of rapamycin (7) and PA stimulates stress fiber formation (8). The relative concentrations of LPA and PA in biological membranes control their curvature and vesicle budding (9). C1P is the sphingolipid analogue of PA and is thought to be involved in synaptic vesicle movement and transport (10). It is formed during neutrophil phagocytosis and it is involved in liposome fusion (11). C1P binds to and activates cytosolic phospholipase A 2 , thereby increasing arachidonate and prostaglandin E 2 production (12). C1P also blocks activation of apoptosis in macrophages by inhibiting acidic sphingomyelinase activity (13).
The lipid phosphate phosphatases (LPPs) are a family of enzymes that de-phosphorylate S1P, LPA, PA, and C1P, thus modulating their signaling (4). Such actions may also generate new signals through the dephosphorylated products sphingosine, diacylglycerol, and ceramide. There are three major isoforms of LPP, each containing six transmembrane spanning domains, an N-glycosylation site, which is not required for activity, and three conserved domains constituting a phosphatase active site (5). When the LPPs are expressed in the plasma membrane, the active site faces the extracellular matrix, thereby allowing LPPs to dephosphorylate external lipid phosphates. This orientation confers the potential to regulate the concentrations of extracellular LPA and S1P and possibly attenuate signaling through their respective receptors (18 -20). Additionally, the extracellular activity of the LPPs promotes the uptake of dephosphorylated products of lipid phosphates, which has been shown to regulate cell movement and survival (21,22). The LPPs are also expressed in intracellular membranes, and they can modify intracellular PA and DAG levels and perturb signaling downstream of G-protein-coupled receptors, including thrombin receptors (6,23). Animal models have demonstrated that LPPs play important roles in regulating development, cell migration, tumor progression, and blood vessel formation (5,22). Although each LPP isoform can have a distinct physiological impact, the specific target lipids and functions of the different isoforms are not well defined. LPP2 has a much more restricted distribution in organs than LPP1 and LPP3. LPP2 is therefore likely to have an isoform-specific biological function in tissues in which it is highly expressed compared with the other isoforms, such as in colon, pancreas, and ovary (24).
The present work arose from our observations that overexpressing LPP2 in fibroblasts produced a very different phenotype of cell proliferation compared with the overexpression of LPP1 or LPP3. Increasing LPP2 activity in rat2 fibroblasts caused a premature entry into S-phase associated with premature cyclin A expression. Conversely, knocking down endogenous LPP2 expression delayed S-phase entry associated with delayed cyclin A expression. The effects of LPP2 required its catalytic activity, and were not mimicked by increasing or decreasing LPP1 or LPP3 activity. Fibroblasts that stably overexpressed LPP2, but not LPP1 or LPP3, eventually arrested in G 2 /M after 20 passages and exhibited changes in the concentration of proteins and lipids that are characteristic of senescence. This work describes a novel, isoform-specific function of LPP2 that regulates cell cycle progression.

EXPERIMENTAL PROCEDURES
Cloning and Expression of LPPs-Rat2 cells and Bosc 31 packaging cells were described previously (18). cDNA for human LPP2, a gift from Dr. A. Morris (University of North Carolina, Chapel Hill, NC), or cDNA for rat LPP3 or mouse LPP1, were subcloned into the pBabePuro (pBP) expression vector. PCR was used to add a GFP tag to the C terminus of LPP2 and to create an R214K mutation. The pBP constructs were transiently transfected into retroviral Bosc 31 packaging cells and virioncontaining media were used to infect rat2 fibroblasts. Mixed populations of transduced cells were selected by puromycin resistance (18). GFP-tagged human LPP2 and myc-tagged mouse LPP1 driven by a cytomegalovirus promoter were transferred into an adenovirus-packing cell line using the AdEasy vector system (Stratagene, La Jolla, CA) following the manufacturer's instructions. The recombinant plasmids were linearized and propagated in HEK 293 cells, and high-titer purified preparations (1 ϫ 10 10 plaque-forming units/ml) were generated by the University of Iowa Gene Transfer Vector Core. For adenoviral transfections, multiplicity of infection 12 plaque-forming units/cell for myc-LPP1 or multiplicity of infection 100 plaque-forming units/cell for LPP2-GFP were added to cells in antibiotic-free media for 24 h.
siRNA Transfection-Double-stranded SMARTpool siRNAs targeting rat LPP1, rat LPP2, rat LPP3, cyclophilin B, and non-targeting controls were purchased from Dharmacon (Lafayette, CO). Lipofectamine 2000 (Invitrogen) in Opti-MEM (Invitrogen) was used at 0.625 g/ml according to the manufacturer's protocol. The final concentration of siRNAs was 200 nM. Controls for the knock-downs were performed with cyclophilin B, non-targeting control siRNAs, and lipofectamine alone. It was determined experimentally that maximum knock-down was achieved at and remained constant between 40 and 72 h posttransfection. The transfection efficiency for the introduction of siRNA was about 90%, as evaluated by the number of fluorescent cells trans-fected with siGLO, divided by the number of nuclei stained with Hoescht 33258 or phase-contrast microscopy (results not shown). For cell cycle analysis, transfection was performed in antibiotic-free media containing serum, and media were changed 6 h after transfection. After a further 18 h of transfection, cells were treated with serum-free media for 20 h before the re-addition of serum to promote cell cycle progression. Lysates were collected for real time RT-PCR at 12 h after the addition of serum in each experiment to determine the extent of knock-down achieved at approximately the point of S-phase entry.
Real-time RT-PCR-RNA was collected using the RNAaqueous kit (Ambion Inc., Austin, TX) according to the manufacturer's directions. Contaminating DNA was removed using the DNA-free kit (Ambion) according to the manufacturer's directions. RNA was quantitated spectrophotometrically at 260 nm. Reverse transcription was performed using Superscript II (Invitrogen), random primers (Invitrogen), and RNAout (Invitrogen) according to the manufacturer's instructions. Negative controls lacking RNA or RT were performed with each reverse transcription reaction. PCR was performed on an Icycler (Bio-Rad). Each reaction contained 0.2 M each primer, ϳ100 ng of cDNA from the reverse transcription reaction, and SYBR Green PCR master mixture (Applied Biosystems, Foster City, CA). Standard curves were generated for each primer pair and the slope and efficiency calculated from the curves were used to determine target RNA levels relative to the housekeeping gene cyclophilin A. Melting curves were performed with each analysis to determine product specificity, and amplified products were run out in 2% agarose to confirm the presence of a single band. An annealing temperature of 57°C was used for all primer pairs. Primers for PCR were as follows: LPP2 forward, TGGCCAAGTA-CATGATTGG and reverse, AGCAGCCGTGCCCACTTCC; LPP1 forward, GGTCAAAAATCAACTGCAG and reverse, TGGCTTGAAG-ATAAAGTGC; LPP3 forward, CCCGGCGCTCAACAACAACC and reverse, TCTCGATGATGAGGAAGGG; and mouse cyclophilin A forward, CACCGTGTTCTTCGACATCAC and reverse, CCAGTGCTCA-GAGCTCGAAAG. Primers for the LPPs were designed to recognize human, mouse, and rat sequences.
Analyses of Proliferation and Apoptosis-Cells were seeded at 30,000 cells/dish and grown for 8 days, with fresh media added each day. Under these conditions, cells proliferated exponentially for 2-3 days before encountering contact inhibition, irrespective of passage number. Cells were washed with HEPES-buffered saline, trypsinized, resuspended in growth media, and counted on a hemocytometer. Parallel determinations of protein and DNA content were performed in some cases using the bicinchoninic acid assay (Bio-Rad) and Hoechst staining in a 96-well plate (26), respectively. For measurement of apoptosis, cells were fixed with buffered 4% formaldehyde and stained with 500 ng/ml Hoechst 33258. Apoptotic cells were quantitated by counting condensed and/or fragmented nuclei versus evenly stained nuclei (27).
Cell Cycle Analysis-Cells were synchronized by starvation in Dulbecco's minimum essential medium containing 0.6% fatty acid-free bovine serum albumin (Sigma) and released after 24 h by adding Dulbecco's minimum essential medium containing 10% FBS. Cells synchronized by trypsinization exhibited the same phenotype (results not shown). Nocodazole and double thymidine block techniques were not used because of their inability to produce adequate cell cycle arrest or re-entry in the rat2 cell line. We used flow cytometry to measure the cell cycle distribution of control and LPP2 overexpressing cells during serum starvation to ensure that the cells were arrested in G 1 to a similar extent. Control and LPP2 overexpressing cells had ϳ70% of cells in G 1 -phase, 20% in S-phase, and 10% in G 2 /M-phase prior to starvation. After 9 h of starvation, both control and LPP2 overexpressing cells contained 85% of cells in G 1 -phase, 5% in S-phase, and 10% in G 2 /M-phase. After 24 h of starvation both cell lines had 94% of cells in G 1 -phase, 2% of cells in S-phase, and 4% of cells in G 2 /M-phase. The cell cycle distribution was maintained for an additional 24 h of starvation in both control and LPP2 overexpressing cells. Additionally, Western blots demonstrated equivalent levels of all cyclins in control and LPP2 overexpressing cells after 24 h of starvation. At specific times after serum stimulation, cells were harvested and suspended at 1 ϫ 10 6 cells/ml in Vindelov's reagent (0.01 M Tris base, 10 mM NaCl, 700 units of RNase I, 7.5 ϫ 10 Ϫ5 M propidium iodide (Sigma), 0.1% Nonidet P-40). Analysis was performed on a FACScan flow cytometer (BD Biosciences) using Cellquest software. A minimum of 20,000 cells were gated based on forward scatter versus side scatter and area versus width to exclude doublets, polyploids, and cell fragments. Modfit Lt. Software (Verity Software House, Inc.) was used to quantitate G 1 , S, and G 2 /M peaks. For determination of apoptosis, cells were fixed in 70% ethanol for 18 h and stained with 100 g/ml propidium iodide. The subdiploid peak was quantitated using Cellquest software (BD Biosciences).
Immunoprecipitation and Cyclin-dependent Kinase-1 (CDK1) Kinase Assay-Lysates from cells overexpressing GFP alone or LPP2-GFP were pre-cleared with protein A-Sepharose beads and incubated with monoclonal anti-GFP (Santa Cruz, B-2) at 1:100 for anti-LPP2 or anti-LPP1 Western blots or with anti-CDK1 (Cell Signaling), at 1:200 for CDK1 kinase measurements. Prior to kinase assay, beads were washed with RIPA buffer (50 mM Tris-HCl, 150 mM NaCl, 50 mM NaF, 2 mM dithiothreitol, 0.1% Triton X-100, 0.1 mM sodium orthovanadate, 10 M leupeptin, 100 g/ml aprotinin, 40 mM ␤-glycerophosphate, and 20 mM p-nitrophenyl phosphate) and then in kinase buffer (40 mM Tris, pH 7.6, 2 mM dithiothreitol, 10 mM MgCl 2 ). Precipitates were incubated in 10 l of kinase buffer containing 1 g of histone H1, 50 pmol of ATP, and 1 Ci of [␥-32 P]ATP for 10 min. Reactions were stopped by adding gel loading buffer and products were separated on SDS-PAGE. Phospho- show mRNA concentrations for untreated parental fibroblasts or cells treated with siRNAs for non-targeting control, rat LPP1, LPP2, or LPP3. Panels C and D show mRNA concentrations for stable cell lines expressing empty vector (pBP), hLPP2, hLPP2-GFP, LPP2(R214K)-GFP, LPP1, or LPP3. mRNA concentrations are normalized to that of the housekeeping gene, cyclophilin. In panels B and D, LPP1 and LPP3 mRNA are shown in black and white columns, respectively. Results are expressed as -fold change compared with rat2 fibroblasts, which is given as 1.
Results are mean Ϯ S.D. from at least four independent experiments. Statistically significant differences (p Ͻ 0.05) from control are indicated by the asterisk. rylated substrate was visualized on a phosphorimager and the bands were cut and quantitated in a scintillation counter.
Lipid Determinations-The LPA assay was performed as described previously (29). For mass spectrometric analysis, methanol extracts were combined with internal standards of 0.5 nmol of each of C 12sphingomyelin, C 12 -ceramide, C 12 -galactosylceramide, C 12 -lactosylceramide, C 20 -sphingosine, C 20 -sphinganine, C 17 -sphingosine 1-phosphate, and C 17 -sphinganine 1-phosphate. Samples were analyzed using liquid chromatography and tandem mass spectrometry (30). To quantitate phosphatidic acid, lipids were extracted using an acidified Bligh and Dyer method and analyzed after loading in the middle of Silica Gel 60 thin layer chromatography plates (31). Plates were developed twice in chloroform:methanol:ammonium hydroxide (65:35:7.5), cut 1 cm above the PA band, turned upside down, and developed in the reverse direction with chloroform:methanol:acetic acid:acetone:water (50:10:10:20:5). PA was visualized with 0.03% Coomassie R-250 in 20% methanol with 100 mM NaCl, or 0.05% primulin in 80% acetone and quantitated by scanning on a Odyssey imager, or a phosphorimager (Bio-Rad) at 525 nm. Diacylglycerol was measured using a DAG kinase assay (31). Results for all lipid analyses were expressed relative to total lipid phosphate (31). For determination of nuclear DAG, nuclei were purified by centrifugation through a 16% sucrose cushion. Nuclei were washed twice with buffer containing 10% sucrose and lipids were extracted as above. The presence of intact nuclei was confirmed by Hoechst staining using a fluorescence microscope.

Characterization of Fibroblasts with Modified Expression of the
LPPs-To study the isoform-specific effects of the LPPs, techniques for decreasing and increasing the relative expression of the three isoforms were developed. Knock-down experiments were performed by transfecting cells with siRNAs for each of the rat LPP isoforms. Real-time RT-PCR demonstrated that rat2 fibroblasts treated with siRNAs for LPP1, LPP2, and LPP3 showed about a 60% decrease in mRNA for the targeted LPP (Fig. 1, A and B). Transfection with control siRNAs did not decrease the expression of any LPP isoforms, and the knock-down of each of the three LPPs did not significantly alter the expression of mRNA for the other isoforms (Fig. 1, A and B).
To overexpress the LPPs, rat2 fibroblasts were transduced with hLPP2, hLPP2-GFP, mutant LPP2(R214K)-GFP, mLPP1, mLPP1-GFP, rLPP3-GFP, or myc-rLPP3 and stable cell populations were selected with puromycin without clonal selection. Cells transduced with LPP2, LPP2-GFP, and R214K-GFP showed 32-, 42-, and 28-fold increases in mRNA for LPP2, respectively, compared with the endogenous expression levels in cells transduced with empty vector (Fig. 1C). Overexpression of LPP1 and LPP3 resulted in 16-and 78-fold increases in mRNA levels, respectively (Fig. 1D). The overexpression of each of the three LPP isoforms did not alter the expression of mRNA for the other isoforms ( Fig. 1, C and D). When RT-PCR reactions were performed using the same reagents and RNA concentrations, the three isoforms had primer efficiencies of 1.88, 1.62, and 1.64, and required 22, 26, and 22 cycles to reach the threshold for LPP1, LPP2, and LPP3, respectively. The higher number of threshold cycles required for LPP2 indicated that LPP2 is likely to be the least abundant isoform in rat2 fibroblasts.
LPP2 protein levels could not be determined because of technical difficulties encountered in resolving the protein on SDS-PAGE. Various techniques, which allowed the resolution of LPP1-GFP and LPP3-GFP with anti-GFP, including the addition of urea, increased detergent concentrations, N-ethylmaleimide addition, and lack of boiling, all failed to resolve LPP2-GFP and untagged LPP2 using two different anti-LPP2 antibodies (23,32) or an anti-GFP antibody. Immunoprecipitation of LPP2-GFP with anti-GFP antibody demonstrated that recombinant LPP2 activity was recovered ( Fig. 2A), and there was no soluble GFP detected on Western blots (results not shown). This indicated that the LPP2-GFP fusion protein was overexpressed and remained intact. The immunoprecipitate from cells stably overexpressing LPP2-GFP did not co-immunoprecipitate LPP1, even when cells were transfected with adenovirus expressing myc-tagged mLPP1 to maximize any possible LPP1-LPP2 interaction (Fig. 2B). This demonstrated that the activity in the immunoprecipitate was caused by the activity of recombinant LPP2-GFP protein, not associated LPP1. When this immunoprecipitate was analyzed by Western blotting, there was a diffuse doublet of ϳ160 -200 kDa that could not be resolved further by any of the techniques described above (results not shown). LPP2 can homodimerize (33), and LPP2 multimerization is a probable cause of the high molecular weight aggregates that could not be resolved. Using adenoviral overexpression of LPP2-GFP, a band at the correct molecular mass of ϳ60 kDa was visualized with anti-LPP2 after immunoprecipitation with anti-GFP (results not shown). This band was only visible at mRNA overexpression levels of 100-fold or more, ϳ2.5 times more than the levels achieved by stable transduction. This result agrees with work of other investigators who have had similar difficulties resolving monomeric LPP2 on SDS-PAGE and have visualized the protein only in conditions of 100fold or greater overexpression (23). Lipid Phosphatase Activity in Transfected Cells-Total lipid phosphatase activity consists of the combined activities of the three LPP isoforms and it was measured in whole cell lysates using PA in Triton X-100 micelles. In cells in which endogenous LPP2 expression was knocked down by 61%, there was no change in total LPP activity (Fig.  2C). However, we are confident that this level of down-regulation of LPP2 mRNA is biologically relevant because it resulted in a clear phenotype (see next section). Knock-down of LPP3 also failed to significantly change lipid phosphatase activity, however, knock-down of LPP1 to 48% of endogenous levels produced a 53% decrease in total LPP activity (Fig. 2C). This suggests that LPP1 is the major contributor to endogenous LPP activity in the fibroblasts. Furthermore, knock-down of LPP1 may be expected to change bulk lipid concentrations in the cells, whereas knock-down of LPP2 or LPP3 would be less likely to do so.
Ecto-LPP activity was measured in intact cells as the dephosphorylation of 10 M LPA or 5 M S1P in the extracellular medium. The overexpression of LPP2 did not significantly change the hydrolysis of extracellular LPA or S1P (results not shown). The overexpression of LPP1 and LPP3 did increase the hydrolysis of extracellular LPA and S1P (results not shown).
Localization of LPP2 in Rat2 Fibroblasts-Confocal studies were performed using antibodies to the GFP tag on LPP2 and to various organelle markers. Wild-type and mutant LPP2 showed the same local-ization profile, which differed from the ubiquitous cellular distribution of GFP alone (supplementary Fig. i). LPP2-GFP and LPP2(R214K)-GFP were localized to the plasma membrane and intracellular membranes. Co-localization studies indicated that LPP2 was found in the early endosomes co-localized with early endosome antigen-1, and co-localized with caveolin-1 at the plasma membrane and in intracellular membranes (supplementary Fig. i). Partial co-localization was observed with the endoplasmic reticulum marker calnexin (results not shown). LPP2 did not co-localize significantly with markers for the Golgi apparatus, mitochondria, nucleus, or nuclear membrane (results not shown). The likely sites of action for LPP2, therefore, include the plasma membrane, endosomes, and endoplasmic reticulum, and it is unlikely that LPP2 acts in the nucleus. Importantly, these results demonstrate that the mutant LPP2 is not mislocalized, and validate using the mutant to distinguish the catalytic versus non-catalytic functions of LPP2.
Decreasing LPP2 Expression Delays S-phase Entry whereas Increasing LPP2 Activity Causes Premature Entry into S-phase-To examine the role of endogenous LPP2 in controlling S-phase entry, we knocked down LPP2 mRNA in rat fibroblasts by an average of 61% in three experiments. We were unable to measure if there was a proportional decrease in LPP2 protein because of the problems described above for Western blotting and lack of antibodies that could detect the low endogenous levels of untagged LPP2. Cells transfected with control siRNAs or   APRIL 7, 2006 • VOLUME 281 • NUMBER 14 rat LPP2 siRNAs were synchronized, and their progression through the cell cycle was measured by flow cytometry. Decreasing endogenous LPP2 expression delayed entry into S-phase by 1.2 Ϯ 0.14 h (mean Ϯ S.D. for three independent experiments) compared with parental control cells, or cells transfected with non-targeting control siRNAs (Fig.  3A). These results confirm that the LPP2 mRNA knock-down produced a physiologically important decrease in LPP2 activity. Furthermore, the effect was specific, because decreasing endogenous LPP1 or LPP3 did not alter the rate of S-phase entry (Fig. 3B). Conversely, LPP2 overexpressing cells entered S-phase 2.4 Ϯ 0.70 h (6 experiments) before control fibroblasts that expressed cDNA for the empty vector, or than those expressing LPP2(R214K) (Fig. 3C). Fibroblasts that overexpressed LPP1 and LPP3 entered S-phase at approximately the same time as vector control cells (Fig. 3D).

Increasing LPP2 Activity Causes Premature Cyclin A Expression and Decreased LPP2 Expression Delays Cyclin A Expression-Levels
of the cyclins that regulate cell cycle progression into S-phase were measured at different times to determine the mechanism of the early S-phase entry. Western blots were quantitated and the results were presented as relative expression levels. These values can be compared within, but not across experiments. Decreasing endogenous LPP2 mRNA delayed cyclin A expression compared with cells treated with control siRNAs (Fig.  4A). Consequently, decreased cyclin A expression occurred between 8 and 14 h after the addition of FBS, prior to S-phase entry. In cells overexpressing catalytically active LPP2, cyclin A expression was accelerated by about 2 h (Fig. 4B). This 2-h acceleration paralleled the 2-h acceleration in S-phase entry. Cells overexpressing LPP1 or LPP3 were indistinguishable from vector control cells in terms of both the timing and magnitude of expression of cyclin A (Fig. 4C). Overexpression of the inactive mutant LPP2(R214K)-GFP did not accelerate the expression of cyclin A (results not shown). LPP2 overexpression did not change the magnitude or timing of expression of cyclins D1, D2, D3, or E, cyclindependent kinase-2, Ser 15 -phosphorylated p53, p21 Cip1 , or p27 (results not shown). Therefore it is probable that LPP2 controls S-phase entry by regulating the timing of cyclin A expression.
Cells Transduced with LPP2 Show Decreased Rates of Proliferation at High Passage and Accumulate in G 2 /M-During our work in culturing cells that overexpressed different LPPs, we consistently observed that the LPP2 overexpressing fibroblasts progressively slowed in their proliferation rates. Cells at passage 24 were seeded at low density and their proliferation was measured for 8 days. After 8 days of growth, control cells and cells overexpressing LPP2(R214K) had increased in number by ϳ40-fold, whereas the numbers of LPP2-transduced cells had increased by only 5-fold (Fig. 5A). Cells transduced with LPP1 or LPP3 proliferated to the same extent as control cells of the same passage (Fig. 5B). The addition of up to 30% fetal bovine serum, 50 M LPA, or 5 M S1P to the media did not overcome the decrease in proliferation exhibited by LPP2-transduced cells (results not shown). The decreased proliferation of LPP2-transduced cells was not caused by increased apoptosis, because both control and LPP2-transduced cells contained only about 1% apoptotic cells, as determined by Hoechst staining or by measuring the subdiploid peak in flow cytometry (results not shown).
To understand the decreased proliferation rate of the LPP2-transduced fibroblasts, we investigated cell cycle progression. After 15-20 passages, cells transduced with LPP2 began to accumulate in G 2 /M. Confluent parental rat2 and vector control fibroblasts at passage 24 contained 85-90% of the cells in G 1 phase and only 4% in G 2 /M, as expected (Fig. 5C). Cells transduced with LPP1, or LPP3, or with inactive mutant LPP2 also contained over 80% of cells in G 1 -phase and less than 8% of cells in G 2 /M-phases at confluence (results not shown). By contrast, at passage 24, about 29% of the cells that were transduced with catalytically active LPP2 were in G 2 /M-phase (Fig. 5C). This number increased with increasing passage number, reaching 70% of cells by passage 35 (results not shown). Hoechst staining confirmed a proportional increase in DNA content per cell in the LPP2-transduced, G 2 -arrested cells (results not shown). At passage 35, cell proliferation became

LPP2 Regulates S-phase Entry
undetectable (results not shown). Cells transfected with the empty vector, or LPP2(R214K) maintained a distribution of greater than 80% of cells in G 1 -phase even after more than 40 passages (results not shown).
Cells Transduced with LPP2 Activate the G 2 /M Checkpoint at High Passage-To investigate if the G 2 /M arrest and decline in proliferation of LPP2-transduced cells was a result of activation at the G 2 /M checkpoint, we measured the phosphorylation state of CDK1. Dephosphorylation of the inhibitory Tyr 15 phosphorylation on CDK1 is required for cells to progress into mitosis. At high passage number, cells transduced with LPP2 showed increases of more than 12-fold in Tyr 15 -phosphorylated CDK1, compared with unsynchronized control cells (Fig. 6A). Cyclin B expression was also decreased by 75% in high passage LPP2-transduced cells, compared with control cells (Fig. 6B). Decreased cyclin B expression is common in cells that have undergone permanent cell cycle exit (34). To test whether the changes in cyclin B expression and CDK1 phosphorylation were the result of checkpoint activation following repeated premature S-phase entry, or were because of a defect in G 2 /M progression caused by LPP2 activity, cyclin B expression and CDK1 phosphorylation were measured in low passage, cycling LPP2 overexpressing cells. In cells prior to passage 20, the phosphorylation of CDK1 on Tyr 15 was similar in vector control and LPP2 overexpressing cells (Fig. 6C). Additionally, in low passage cells the expression of cyclin B was similar in vector control and LPP2 overexpressing cells (Fig. 6D). The peak Tyr 15 phosphorylation of CDK1 occurred 2 h earlier in LPP2 overexpressing cells, because of the premature S-phase entry that led to 2 h earlier entry into mitosis (Fig. 6C). Because cyclin B levels were not decreased in LPP2 overexpressing cells at low passage number, and because these cells showed normal phosphorylation of CDK1, it is likely that activation of the G 2 /M checkpoint at high passages is not a direct result of LPP2 activity. In cells entering S-phase prematurely, and presumably unchecked, randomly occurring DNA damage would not be repaired because of lack of time in G 1 . Thus, over time, unrepaired DNA damage could accumulate in cells with persistent unscheduled S-phase entry.
Cells Transduced with LPP2 Show Characteristics of Senescence at High Passage-LPP2-transduced cell populations at late passage number, in which greater than 30% of cells were arrested in G 2 , displayed many changes in protein expression that are characteristic of DNA damage or senescence. The level of phospho-p53 (Ser 15 ) was elevated 16-fold, and expression of p21 Cip1 , p27, and p16 were increased by 8-, 6-, and 7-fold, respectively (Fig. 6E). Additionally, cyclins D1, D2, D3, and E were increased 5-, 7-, 2-, and 4-fold, respectively (Fig. 6E). These increases in cyclin expression are consistent with previous studies in which cyclin D and E levels were elevated in senescent cells (35,36). Surprisingly, LPP2-transduced cells containing more than 50% of cells in G 2 /M with an activated G 2 /M checkpoint also eliminated the overexpression of LPP2, as determined by real-time RT-PCR (results not shown). At passage 35, LPP2 mRNA levels were not statistically different from LPP2 mRNA expression levels in rat2 control cells.
Cells arrested in G 2 were also analyzed for lipid content. G 2 /M-arrested cells contained more than twice the relative amount of ceramide of parental cells (Table 1). G 2 -arrested cells also showed a 50% decrease in LPA levels relative to total phospholipid (Table 1). Sphinganine phosphate levels also appeared to have increased in G 2 -arrested cells, but the effect was not statistically significant. The changes observed in ceramide and LPA concentrations were not observed in cycling LPP2 overexpressing cells at early passages, and are therefore related to the G 2 /M arrest phenotype. Other lipids measured including ceramide 1-phosphate, sphingosine, sphingosine 1-phosphate, and sphinganine were not changed significantly in G 2 -arrested cells compared with control cells (Table 1). Phosphatidate and total and nuclear diacylglycerol levels were also not significantly different in LPP2 overexpressing or in G 2 -arrested cells compared with control fibroblasts (results not shown).

DISCUSSION
Little is known about the specificity and functions of the different LPP isoforms and how they differentially modify cell signaling. In this study we demonstrate that LPP2 regulates cell cycle progression. Decreasing the expression of endogenous LPP2 delays S-phase entry, whereas increasing LPP2 expression results in premature entry into S-phase. LPP2 catalytic activity was required for these effects because expression of the inactive LPP2(R214K) mutant did not change the rate of S-phase FIGURE 6. Cells that stably overexpressed LPP2 and are arrested in G 2 show activation of the G 2 /M checkpoint and characteristics of senescence. Panels A and B show the quantitation of Western blots for phospho-CDK1 (Tyr 15 ) and cyclin B, respectively, in cells at passage 26 that were stably overexpressing empty vector or LPP2. Expression is relative to the vector control, which is given as 1. Upper panels show images of the membranes as scanned by the Odysseyா imager. Panels C and D show the quantitation of Western blots for phospho-CDK1 (Tyr 15 ) and cyclin B, respectively, in cells stably transduced with empty vector (Vector, f) or LPP2 (LPP2, F), at low passage, at various times after synchronization and the addition of FBS. Expression is relative to the empty vector at time 0, which is given as 1. Results are from one representative of three independent experiments. Panel E shows quantitations of Western blots for cyclin D1, cyclin D2, cyclin D3, cyclin E, phospho-p53 (Ser 15 ), p21 Cip1 , p27, and p16 in cells transduced with LPP2 and grown asynchronously at passage 26. Expression is shown relative to expression in cells transduced with empty vector and grown asynchronously at passage 26, which is given as 1. Results are mean Ϯ S.D. from at least three independent experiments. Statistically significant differences (p Ͻ 0.05) from vector control are indicated by the asterisk. APRIL 7, 2006 • VOLUME 281 • NUMBER 14 entry. The use of the inactive mutant is justified because the mRNA expression and subcellular distribution of the protein were not significantly different from wild-type LPP2. The effects of LPP2 were isoform specific, because increasing or decreasing the expression of LPP1 or LPP3 using the same protocols did not alter the rate of S-phase entry. The profound effects produced by knocking down endogenous LPP2 activity illustrates that LPP2 is an important regulator of S-phase entry. This work, therefore, provides the first evidence of an isoform-specific biological function for LPP2 activity in regulating cell cycle progression.

LPP2 Regulates S-phase Entry
Overexpression of catalytically active LPP2 resulted in premature S-phase entry after synchronization by serum deprivation. We ensured that this did not result from inadequate arrest in G 1 during serum deprivation (see "Experimental Procedures"). To ensure that overexpression of LPP2 reproducibly and selectively accelerated S-phase entry, we transduced rat2 fibroblasts by retroviral infection with human LPP2 or human LPP2 tagged at the C terminus with GFP. Polyclonal cell populations were used and LPP2 was subcloned into both the pBabePuro and pLNCX2 vectors, which have different selection markers. Stable cell populations transduced with empty vector, untagged LPP2, LPP2-GFP, LPP2(R214K)-GFP, LPP1, and LPP3 were created on four separate occasions. Every cell population created that overexpressed catalytically active tagged or untagged LPP2 entered S-phase prematurely. These results establish that the GFP tag on LPP2 did not change its effect on cell cycle regulation. By contrast, every cell population expressing the empty vector control, LPP1, LPP3, or mutant LPP2(R214K) did not show changes in the timing of entry into S-phase.
The effect of LPP2 on S-phase entry appears to be regulated through cyclin A. The increase in cyclin A expression was accelerated by about 2 h in cells overexpressing LPP2 activity and it was delayed by about 1.5 h in cells with decreased LPP2 expression. These changes corresponded to the acceleration, or delay in S-phase entry. The changes in cyclin A also required the catalytic activity of LPP2, and cyclin A expression was not changed by modulating the activities of LPP1 or LPP3. Cyclin A is a partner of cyclin-dependent kinase-2 (CDK2), which regulates G 1 -to S-phase progression. Dysregulation of cyclin A expression and subsequent increases in cyclin A-associated CDK2 activity leads to unscheduled progression into S-phase (37)(38)(39)(40)(41)(42)(43)(44). The expression of other cell cycle regulatory proteins (p21 and p27, cyclins D1, D2, D3, and E) were unchanged in LPP2 overexpressing cells that entered S-phase prematurely. Furthermore, differences in cyclin A expression occurred at time points prior to S-phase entry. We, therefore, conclude that LPP2 mediates its effects on S-phase entry primarily through regulating the timing of cyclin A expression. Several kinases that influence cyclin A expression and G 1 to S-phase progression (ERK, p38 MAPK, Akt, and LIM kinase) were not changed in expression level, timing of expression, or phosphorylation state in cells that overexpressed LPP2 and entered S-phase prematurely (results not shown). We, therefore, conclude that LPP2 does not increase cyclin A expression through ERK, p38 MAPK, Akt, or LIMK. To determine whether LPP2 expression is itself regulated during cell cycle progression, we measured endogenous levels of LPP2 mRNA in rat2 fibroblasts during starvation and throughout the 24 h of the cell cycle following stimulation with serum. The level of mRNA for LPP2 remained constant during starvation and during cell cycle progression (results not shown). These results do not exclude the regulation of LPP2 activity by post-translational modification or subcellular localization to control the rate of S-phase entry in relevant physiological situations.
Our results from real time RT-PCR and changing LPP2 mRNA expression indicate that LPP2 is not a major contributor to the overall LPP activity or ecto-LPP activity in fibroblasts. Therefore, it is not surprising that we were unable to identify a change in the bulk concentration of a bioactive lipid (PA, DAG, LPA, or ceramide) that would explain the regulation of S-phase entry. It is likely that it is the regulation of a specific pool of bioactive lipid, not the bulk concentration, that is responsible for LPP2-induced changes in the timing of cyclin A expression and S-phase entry. It was technically impractical to obtain enough cells to separate cell fractions and determine the subcellular concentrations of low abundance lipids at multiple time points during the cell cycle. Even if this could be achieved, it is doubtful that lipids like LPA and S1P would remain associated with the original organelle during fractionation. We did determine nuclear levels of DAG (a potential product from LPP2 action on PA) at 2-h intervals from the point of release from starvation until mitosis in cycling cells, and no significant effect of LPP2 was observed. This result is not surprising because confocal microscopy demonstrated that a large portion of LPP2 is present in membranes of early endosomes with some in the endoplasmic reticulum, and that it is absent from the nuclear membrane. It is, therefore, predicted that lipid pools in these former organelles are the most likely to have been affected initially by LPP2. It is not uncommon to observe biological consequences of LPP activity that cannot easily be attributed to specific changes in lipid concentrations (22,45). The substantial changes in cell cycle progression produced by changing LPP2 activity demonstrate that despite the low endogenous expression of LPP2 compared with the other isoforms, LPP2 activity can regulate cell signaling in fibroblasts.
We also determined the long-term effects of LPP2 overexpression. Cells that overexpressed catalytically active LPP2 began to accumulate in G 2 /M between 15 and 35 passages. These cells eventually exited the cell cycle and showed permanent G 2 arrest. The fact that every cell line that overexpressed inactive LPP2(R214K), LPP1, or LPP3 continued to cycle and never showed G 2 /M arrest, even after more than 40 passages, demonstrates that the arrest is specific to the phenotype produced by LPP2. Cell populations with greater than 50% of cells in G 2 /M eliminated the overexpression of LPP2. The suppression of LPP2 activity could have been necessary to permit cells to maintain G 2 arrest and cease cycling. Cell populations containing more than 30% of cells in G 2

TABLE 1 Lipid composition of G 2 arrested cells
Concentrations of bulk cellular sphingolipids were determined by mass spectrometry in parental fibroblasts (R2) and cells showing the G 2 arrest phenotype subsequent to LPP2 overexpression. Concentrations are expressed relative to total sphingomyelin. Samples were analyzed in triplicate and results are expressed as mean Ϯ S.D. for at least three independent determinations. LPA concentrations were determined in cells overexpressing empty vector, low passage cells overexpressing LPP2, or cells formerly overexpressing LPP2 that were arrested in G 2 . LPA concentration is normalized to total phospholipid, and expressed as -fold increase where the vector control is 1. Results are mean Ϯ S.D. for three independent experiments.

Cells
Relative

LPP2 Regulates S-phase Entry
had markedly increased levels of cyclins D1, D2, D3, and E, phosphorylated p53 (Ser 15 ), p21, p27, and p16 INK4a , characteristic of a G 2 -arrested or senescent phenotype (35,36,46,47). In these cell populations, cycling was virtually undetectable and cyclin levels did not vary over time, even after cells were starved by serum deprivation. Cyclin B levels were reduced compared with unsynchronized control cells, and cyclin A levels were similar to the levels in control cells. The level of inhibitory phosphorylation of Tyr 15 on CDK1 in G 2 -arrested cells was similar to that at its maximal activation prior to the G 2 /M transition in cycling control cells, and remained constitutively at this level. Increased Tyr 15 phosphorylation of CDK1 is commonly observed in cells with DNA damage. The G 2 /M checkpoint activation in late passage cells transduced with LPP2 likely resulted from accumulation of DNA damage resulting from repeated premature S-phase entry because LPP2 overexpressing cells at low passage showed normal expression of cyclin B and normal regulation of CDK1 phosphorylation. In cultured cells, some oncogenes can induce premature senescence after initially stimulating proliferation, and this process may represent a physiological response involved in preventing malignancy (48,49). This type of senescence is characterized by the up-regulation of p53 and p16 INK4a (48). Fibroblast populations that were largely arrested in G 2 as a consequence of initial LPP2 overexpression contained about twice as much ceramide as control cells. Different ceramide species were increased proportionally, and the predominant species, 16:0, comprised 50% of the total ceramide. Ceramide levels increase in senescent cells and increased sphingomyelinase activity and high ceramide concentrations are instrumental in maintaining a senescent phenotype (50,51). The G 2 -arrested cells also had significantly lower levels of LPA than control cells. To our knowledge, this is a novel finding, and could suggest a previously unknown role for LPA in growth regulation and senescence. LPA is an agonist for the peroxisome proliferator-activated receptor-␥ receptor (14), which decreases the synthesis of several proteins that are increased in senescence, including cyclin D, cyclin E, p21, and p27 (52). Therefore, it is possible that decreased LPA and decreased peroxisome proliferator-activated receptor-␥ signaling could contribute to the high expression of these proteins and the senescent phenotype. Concentrations of other cellular lipids, including ceramide 1-phosphate, sphingosine, sphingosine 1-phosphate, and sphinganine were not significantly changed in cell populations that were arrested in G 2 . It is important to note that the changes observed in ceramide and LPA concentrations were seen in cells in which the overexpression of LPP activity had been overcome. Therefore, these changes relate to the senescent phenotype and G 2 arrest.
Our results indicate that LPP2 regulates timing of entry into S-phase, but it is not essential for cell-cycle progression. Several genes that regulate progression into late G 1 or entry into S-phase have been knocked out in mice without lethality or other major generalized phenotypes. These knockouts include critical cell-cycle regulators such as CDK2, CDK4, CDK6, and cyclins D1, D2, D3, E1, or E2 (reviewed in Ref. 53). Therefore, deletion of LPP2 would not be expected to result in lethality or any other major generalized phenotype. Consistent with this expectation, LPP2 knock-out mice are viable and overtly normal (54). By contrast, knocking out LPP3 expression causes embryonic lethality (55). Transgenic mice that overexpress LPP1 have decreased birth weight, sparse curly hair, and defective spermatogenesis causing infertility (45). Therefore, these studies with mouse models support our work demonstrating that LPP2 has a unique and isoform-specific function that is not exhibited by LPP1 and LPP3. Our studies show that this unique function is the regulation of the timing of entry into S-phase.
In summary, this study demonstrates that LPP2 is a regulator of cell cycle progression in fibroblasts. Decreasing the expression of LPP2 caused a 1.5-h delay in entry into S-phase following the delayed expression of cyclin A. Overexpression of LPP2 caused the premature expression of cyclin A and a 2-h premature entry into S-phase. These represent substantial changes in the rate of S-phase entry that could have implications in processes such as mitogenesis, migration, wound healing, development, and tumorigenesis. Cell cycle regulation depended on the catalytic activity of LPP2, and this effect was isoform specific. Overexpression or knock-down of LPP1 or LPP3 did not alter S-phase entry. Cells that overexpressed catalytically active LPP2, but not inactive LPP2, LPP1, or LPP3, accumulated in G 2 /M-phase of the cell cycle progressively after 20 passages as a result of activating the G 2 /M checkpoint. These cells eventually stopped proliferating and exhibited changes in protein and lipid concentrations characteristic of DNA damage and senescence. This work provides the first evidence of a catalytic and isoform-specific function of LPP2 as a cell cycle regulator.