Autotaxin Is Released from Adipocytes, Catalyzes Lysophosphatidic Acid Synthesis, and Activates Preadipocyte Proliferation

Our group has recently demonstrated (Gesta, S., Simon, M., Rey, A., Sibrac, D., Girard, A., Lafontan, M., Valet, P., and Saulnier-Blache, J. S. (2002) J. Lipid Res. 43, 904–910) the presence, in adipocyte conditioned-medium, of a soluble lysophospholipased-activity (LPLDact) involved in synthesis of the bioactive phospholipid lysophosphatidic acid (LPA). In the present report, LPLDact was purified from 3T3F442A adipocyte-conditioned medium and identified as the type II ecto-nucleotide pyrophosphatase phosphodiesterase, autotaxin (ATX). A unique ATX cDNA was cloned from 3T3F442A adipocytes, and its recombinant expression in COS-7 cells led to extracellular release of LPLDact. ATX mRNA expression was highly up-regulated during adipocyte differentiation of 3T3F442A-preadipocytes. This up-regulation was paralleled by the ability of newly differentiated adipocytes to release LPLDact and LPA. Differentiation-dependent up-regulation of ATX expression was also observed in a primary culture of mouse preadipocytes. Treatment of 3T3F442A-preadipocytes with concentrated conditioned medium from ATX-expressing COS-7 cells led to an increase in cell number as compared with concentrated conditioned medium from ATX non-expressing COS-7 cells. The specific effect of ATX on preadipocyte proliferation was completely suppressed by co-treatment with a LPA-hydrolyzing phospholipase, phospholipase B. Finally, ATX expression was found in mature adipocytes isolated from mouse adipose tissue and was substantially increased in genetically obese-diabeticdb/db mice when compared with their lean siblings. In conclusion, the present work shows that ATX is responsible for the LPLDact released by adipocytes and exerts a paracrine control on preadipocyte growth via an LPA-dependent mechanism. Up-regulations of ATX expression with adipocyte differentiation and genetic obesity suggest a possible involvement of this released protein in the development of adipose tissue and obesity-associated pathologies.

adipose tissue plays a crucial role in energy balance. In obesity, excessive accumulation of triacylglycerol in adipocytes (hypertrophy) results from an alteration in the balance between lipogenic and/or lipolytic activities of the adipocytes.
It is now recognized that, beside their involvement in lipid homeostasis, adipocytes also produce and secrete numerous factors. Among them are endocrine peptides (leptin, adiponectin, angiotensinogen, etc.) which may play an important role in the development of morbid complications of obesity such as cardiovascular diseases, hypertension, diabetes, and cancer. Other adipocyte-secreted factors (tumor necrosis factor, fatty acids, eicosanoids, lysophosphatidic acid, etc.) are produced locally and may influence adipose tissue development and/or metabolism by exerting autocrine/paracrine effects on the different cells composing adipose tissue (adipocytes, preadipocytes, and endothelial cells) (1)(2)(3). Of particular interest is the ability of some adipocyte-secreted factors to exert a paracrine control on preadipocyte proliferation and differentiation, cellular processes leading to the recruitment of new fat cells in adipose tissue (adipogenesis) and a further increase in adipose tissue mass (3,4).
Our group has demonstrated that lysophosphatidic acid (LPA) 1 is released from adipocytes in vitro and is present in vivo in the extracellular fluid of adipose tissue collected by microdialysis (5). In parallel, LPA is able to activate preadipocyte motility and proliferation by interacting preferentially with the LPA1 receptor (LPA1-R) (6). Therefore LPA may participate to the paracrine control of adipogenesis.
LPA is a bioactive phospholipid regulating a wide range of cellular responses (proliferation, survival, motility, ion flux, and secretion) through the activation of the G-protein-coupled receptors LPA1-, LPA2-, and LPA3-R (7,8). Bioactive LPA was initially found in culture serum (9) and was further detected in other biological fluids such as plasma (10,11), ascitic fluid (12), follicular fluid (13), aqueous humor (14), and the extracellular fluid of adipose tissue (5). Whereas serum LPA mainly originates from aggregating platelets (15,16), the precise cellular origin of LPA in other biological fluids still remains unclear.
We recently showed that, in parallel to LPA, adipocytes also release a LPA-synthesizing activity that was characterized as a lysophospholipase D activity (LPLDact), catalyzing transformation of lysophosphatidylcholine into LPA (17). Adipocyte LPL Dact is soluble and sensitive to cobalt ions (17). In addition, adipocyte lysophospholipase D activity is insensitive to primary alcohol, suggesting that it could be catalyzed by a non-conventional phospholipase D that remains to be identified.
In the present study, adipocyte LPLD activity was purified and identified as the type II ecto-nucleotide pyrophosphate phosphodiesterase autotaxin (ATX). ATX was found to be expressed by and released from adipocytes, activates preadipocyte proliferation, and its expression was strongly up-regulated during adipocyte differentiation and in a model of genetic obesity. Because of its LPLDact, adipocyte-ATX leads to LPA synthesis and could be involved in the control of adipose tissue development as well as in obesity-associated pathologies.

EXPERIMENTAL PROCEDURES
Culture and Transfection Studies-The mouse preadipose cell line 3T3F442A (18) was grown and differentiated as described previously (19). Briefly, cells were grown to confluence (day 0 of differentiation) in DMEM supplemented with 10% donor calf serum and then shifted in a differentiating medium consisting in DMEM supplemented with 10% fetal calf serum plus 50 nM insulin. In these culture conditions, quiescent preadipocytes differentiate into functional adipocytes. COS-7 monkey cells (American Type Culture Collection) were grown in DMEM supplemented with 10% fetal calf serum and transfected using DEAEdextran as reported previously (20).
Preparation of Conditioned Media-3T3F442A cells or COS-7 cells were washed twice with phosphate-buffered saline to remove serum and incubated (5 ml for a 10-cm diameter plate; 1 ml for a 3-cm diameter plate) in serum-free DMEM supplemented (for LPA release) or not (for LPLDact release) with 1% free fatty acid BSA at 37°C in a humidified atmosphere containing 7% CO 2 . Conditioned media were also prepared from adipose tissue by incubating 300 -500 mg of finely cut out subcutaneous adipose tissue from db/ϩ or db/db mice in 3 ml of serum-free DMEM to measure released LPLDact. After various time of incubation (0 -5 h for measurement of LPLD activity; 7 h for quantification of LPA and lysophosphatidylcholine (LPC)), conditioned medium (CM) was separated from the cells or tissue, centrifuged to eliminate cell debris, and stored at Ϫ20°C before measurement of LPLDact or quantification of LPA. In some experiments, conditioned media were concentrated (about 50-fold) using an Amicon Ultra 10,000 (Millipore). After adjustment of the protein concentration, concentrated conditioned media (CCM) were aliquoted and stored at Ϫ20°C before use.
Measurement of Lysophospholipase D Activity-LPLDact was measured by conversion of radiolabeled LPC into radiolabeled LPA as described previously (21) with minor modifications. A solution of [ 14 C]palmitoyl-lysophosphatidylcholine (PerkinElmer Life Sciences; 55.8 mCi/mmol) at 0.0025 Ci/l in DMEM supplemented with 1% free fatty acid BSA was first prepared, and 20 l of this solution was incubated with 500 l of thawed CM plus 1 l of sodium orthovanadate (0.5 mM) for 90 min at 37°C. At the end of the incubation period, phospholipids were extracted with 500 l of 1-butanol, evaporated, spotted on a silica gel 60 TLC glass plate (Merck), and separated using CHCl 3 /MeOH/NH 4 OH (60: 35:8) as the migration solvent. The plate was autoradiographed overnight at Ϫ80°C using a Biomax-MS film (Kodak) to localize radiolabeled LPA spots, which were scraped and counted with 3 ml of scintillation mixture.
Purification of Lysophospholipase D Activity-All the procedures were performed at 0 -4°C. One hundred twenty milliliters of 3T3F442A adipocyte-conditioned medium containing LPLDact was centrifuged for 1 h at 100,000 ϫ g to remove cell debris and concentrated 8-fold on polyethylene glycol 20,000 (PEG 20,000). After overnight dialysis against 10 liters of purification buffer (PB) using Spectra-Por 1.7 ml/cm tubing (Pierce Chemicals, Interchim) to remove salts and small proteins, 16 ml of the concentrated medium was applied (60 ml/h) onto a heparin-Sepharose column (CLϩ6B, Amersham Biosciences; 15-ml packed volume) equilibrated with 100 ml of PB. The column was washed with 150 ml of PB and eluted with 125 ml of PB containing 0.5 M NaCl. The elution fraction containing LPLDact was dialyzed overnight against 10 liters of PB using Spectra-Por 1.7 ml/cm tubing (Pierce Chemicals, Interchim) to remove NaCl and concentrated on PEG 20,000 to obtain a volume of 40 ml. This concentrate was applied (60 ml/h) on a phosphocellulose column (P11, Whatman, 15 ml packed volume) equilibrated with 100 ml PB. The column was washed with 150 ml of PB and eluted with 125 ml of PB containing 0.5 M NaCl. The elution fraction containing LPLDact was dialyzed overnight against 10 liters of purification buffer using Spectra-Por 1.7 ml/cm tubing (Pierce Chemicals, Interchim) to remove NaCl, and concentrated on PEG 20,000 to obtain a final volume of 5 ml. This concentrate was separated (30 ml/h) over a gel filtration column (HiLoad 16/60, Superdex 200 PrepGrade, Amersham Biosciences) on an Amersham Biosciences FPLC system previously equilibrated with PB. The column was eluted with the same buffer. Aliquots of the collected fractions were assayed for LPLDact. All the fractions containing LPLDact were pooled and concentrated 10 times on dialysis tubing (0.5 ml/cm) in PEG 20,000. The protein concentration was determined by the Bradford assay (Protassay, Bio-Rad) with bovine serum albumin as standard.
SDS-PAGE Separation-SDS-PAGE (4 -20%) was performed according to Laemmli (22) followed by SYPRO Ruby or colloidal blue staining. After the addition of sample buffer (Novex, Invitrogen), concentrated fractions from gel filtration were boiled at 100°C for 5 min. Electrophoretic separation of proteins was carried out on a 1-mm thick 8 ϫ 6-cm gel 10% acrylamide. A 40-g portion of total protein in sample buffer was loaded into a 4-mm well of the gel and separated at 40 mA. A total of 30 g of standards (Mark12, Invitrogen) migrated in a neighboring lane. After coloration with colloidal Coomassie Blue (Biosafe, Bio-Rad), the 110-kDa protein was cut, reduced, and alkylated using dithiothreitol and iodoacetamide, respectively, and subjected to digestion with trypsin overnight following the protocol of Shevchenko et al. (23). Gel pieces were successively washed with ammonium bicarbonate and dehydrated with acetonitrile. After drying in a Speedvac (Heto), the pieces were incubated at 56°C with 10 mM dithiothreitol and then at room temperature with 55 mM iodoacetamide. After successive washing steps with ammonium bicarbonate and acetonitrile, the gel pieces were completely dried in the Speedvac. Trypsin solution (Promega) was added, and digestion was performed overnight at 37°C. Peptide digests were extracted successively with acetonitrile and 5% formic acid. Before analysis, the peptides were dried in the Speedvac, diluted in 5% formic acid, and desalted on Poros R-2 resin. Peptides were eluted with 2 l of 60% methanol/1% formic acid directly into nanospray capillary needles (Micromass, Manchester, UK).
Tandem Mass Spectrometry-Peptides were analyzed on a Q-TOF 2 mass spectrometer (Micromass). MS acquisitions were performed within the mass range of 550 -1300 m/z and MS/MS within 50 -2000 m/z. Data Base Searching and Sequence Analysis-Amino acid sequences were used to search Swiss-Prot and TREMBL with BLAST interface (United States National Center for Biotechnology Information). Partial amino acid sequences (sequence TAG) were used to search the data bases with PeptideSearch interface (European Molecular Biology Laboratory, Heidelberg, Germany). Hits were confirmed by performing a theoretical trypsin digestion with GPMAW (Lighthouse data, Odense, Denmark) and comparing the resulting peptide mass and sequence to those obtained experimentally.
Real Time RT-PCR Analysis-Total RNAs were isolated using the RNeasy mini kit (Qiagen). Total RNAs (1 g) were reverse transcribed for 60 min at 37°C using Superscript II reverse transcriptase (Invitrogen) in the presence of a random hexamer. A minus RT reaction was performed in parallel to ensure the absence of genomic DNA contamination. Real time RT-PCR was performed starting with 25 ng cDNA and 900 nM (human ATX and mouse ATX) or 900 nM (mouse ap2) concentration of both sense and antisense primers in a final volume of 25 l using the SYBR green TaqMan Universal PCR Master Mix (Applied Biosystems). Fluorescence was monitored and analyzed in a GeneAmp 5700 detection system instrument (Applied Biosystems). Analysis of the 18 S ribosomal RNA was performed in parallel using the ribosomal RNA control Taqman Assay Kit (Applied Biosystem) to normalize gene expression. Results are expressed as 2 (Ct18S Ϫ Ctgene) , where Ct corresponds to the number of cycles needed to generate a fluorescent signal above a predefined threshold.
Cloning of Autotaxin from 3T3F442A Adipocytes-Total RNA from 15 days post-confluent 3T3F442A adipocytes was reverse transcribed using random hexamers and Superscript II reverse transcriptase (Invitrogen). First strand cDNA (corresponding to 1 g of total RNA) was amplified using a program consisting of 30 cycles of 95°C for 30 s, 55°C for 30 s, and 72°C for 3 min with a pre-and post-incubation of 95°C for 2 min and 72 for 10 min, respectively. PCR amplification utilized oligonucleotide primers based on the GenBank TM entry for mouse autotaxin (forward 59 -64, reverse 2630 -2651, accession number BC003264). The PCR fragment was isolated and ligated into pcDNA3 (Invitrogen) downstream of the FLAG peptide sequence (Sigma). The recombinant plasmid, designated pcDNA-mATX-FLAG, was sequenced on both strands by automated sequencing.
Quantification of LPA and LPC-LPA was butanol-extracted from conditioned medium and quantified using a radioenzymatic assay as described previously (11). The amount of lysophosphatidylcholine present in conditioned media was determined as described previously (17) after 30 min of treatment with bacterial phospholipase D followed by LPA quantification.
Housing and Treatment of Animals-Male C57BL/KsJ db/db and db/ϩ mice (Jackson Laboratory) were housed in an animal room maintained at 22°C on a 12:12 h light/dark cycle with ad libitum access to food and water. Animals were handled in accordance with the principles and guidelines established by the National Institute of Medical Research (INSERM). On the day of sacrifice, mouse blood was collected on heparin and glucose was immediately measured with a glucose meter (Glucometer 4; Bayer Diagnosis, Puteaux, France). Perigonadic and inguinal white adipose tissues were removed, and adipocytes were isolated as described above and immediately processed for RNA extraction using an RNeasy mini kit (Qiagen). The other tissues (interscapular brown adipose tissue, brain, kidney, and liver) were removed and snap frozen in liquid nitrogen before RNA isolation using RNA STAT kit (AMS Biotechnology Ltd., Oxon, UK).

Isolation and Culture of Preadipocytes and Adipocytes from Mouse
Adipose Tissue-Adipose tissue was dissected from mice and digested in 5 ml of DMEM supplemented with 1 mg/ml collagenase, 1% BSA, and 2 g/ml gentamycin for 45-60 min at 37°C under shaking. Digestion was followed by filtration through a 100-m screen and centrifugation at 800 ϫ g for 10 min at room temperature. This phase allowed the separation of floating adipocytes from the pellet containing the stromalvascular fraction. Adipocytes were washed twice in DMEM and further processed for RNA extraction using the RNeasy mini kit (Qiagen). The stromal-vascular pellet was washed twice in DMEM and resuspended in 1 ml of lysis buffer (7 mM Tris, 0.83% NH 4 Cl, pH 7.6) for 3 min to get rid of erythrocytes and then seeded in 12-well plates (150,000 cells per plate) in DMEM/Ham's F-12 medium supplemented with 10% fetal calf serum. After overnight culture, the medium was replaced by a serumfree differentiating medium consisting of DMEM/Ham's F12 supplemented with transferrin (10 g/ml), biotin (33 M), insulin (66 nM), T3 (1 nM), and pantothenate (17 M). That time point is considered as day 0. In these conditions, adipocyte differentiation occurs a few days later (24). Total RNAs were prepared at different time points after induction of the differentiation using the RNeasy mini kit (Qiagen).

Co-purification of Adipocyte Lysophospholipase D Activity
with Autotaxin-To identify the enzyme responsible for LPL-Dact present in adipocyte CM (17), its purification from 3T3F442A adipocyte CM was undertaken. LPLDact hydrolyzes the phosphocholine bound in LPC to generate LPA and free choline. In the present work, LPLDact was measured using [ 14 C]lysopalmitoyl-phosphatidylcholine as substrate, followed by TLC separation of synthesized [ 14 C]LPA (see "Experimental Procedures"). The initial release of LPLDact in adipocyte CM was observed in an Hepes buffer incubation medium (17). As shown in Fig. 1, a stronger (3.3-fold after 5 h of incubation) LPLDact was measured when using DMEM as the incubation medium. In both incubation media, LPLDact was detectable after a 30-min incubation and reached a maximum after 5 h. Based upon these results, purification of LPLDact was undertaken from 3T3F442A adipocyte CM prepared after 5 h of incubation in DMEM.
After 10-fold concentration, CM was first applied onto a heparin affinity chromatography column (Fig. 2, section 1), washed and eluted with 0.5 M NaCl at once (lane e), and concentrated again. This allowed the recovery of LPLDact, but no specific band was identified by SDS-PAGE (lane f). As a second purification step, concentrated fraction eluted from the heparin column was applied onto a phosphocellulose column (Fig. 2), eluted at 0.2 M NaCl (lane i), concentrated 25 times (lane j), applied on a gel filtration column (Fig. 2, section 3), and eluted (Fig. 2, section 3, fractions 52 to 82).  (Table I) with mouse ecto-nucleotide pyrophosphatase phosphodiesterase (ENPP2), also called autotaxin (25,26). These results showed that adipocyte-LPLDact was co-purified with ATX.
Cloning of 3T3F442A Adipocyte Autotaxin-A unique cDNA encoding ATX was cloned from 3T3F442A adipocytes and subcloned in a pcDNA-FLAG vector. This cDNA exhibited 100% identity with previously identified mouse-ATX cDNA present in the GenBank TM data base (BC003264, NM_015744, AF123542) ( Table I). The corresponding protein is composed of 863 amino acids (molecular mass of ϳ100 kDa) which, by analogy with human, corresponds to the previously described ␤-form (27)(28)(29). In human ATX the site of cleavage has been located between Ser-48 and Asp-49 in the amino acid sequence EGPPTVLS2DSPWTN, which is entirely conserved in mouse (BC003264, NM_015744, AF123542, and herein the 3T3F442A-cDNA sequence). Based upon mouse ATX sequence, the expected cleavage site of mouse ATX should be between Ser-47 and Asp-48. This should lead to a secreted protein of 816 amino acids (molecular mass of ϳ94 kDa). Purification of LPLD-activity from 3T3F442A adipocytes led to identification of a protein with a molecular mass between 97 and 116 kDa (Fig. 1). This mass is slightly higher than that calculated for the secreted form of mouse-ATX. Such a discrepancy could be attributed to the presence, in the mouse sequence, of at least four glycosylation sites (Net Nglyc 1.0 prediction analysis) that would increase the apparent molecular weight of the secreted protein.
ATX Behaves as a Secreted Lysophospholipase D-To determine whether ATX indeed behaved as a secreted lysophos-pholipase D, 3T3F442A adipocyte ATX cDNA was transiently transfected in COS-7 cells. After transfection with empty pcDNA vector, COS-7 cells did not release LPLDact in culture medium (Fig. 3). Conversely, transfection with pcDNA-mATX-FLAG vector led to the release of LPLDact in the culture medium (Fig. 3). Similar results were obtained when transfecting COS-7 cells with a pcDNA vector expressing human ATX (30,31).
Up-regulation of ATX with Adipocyte Differentiation-We previously observed that adipocytes release more LPLDact than preadipocytes (17). We therefore tested the existence of an adipocyte differentiation-dependent regulation of ATX. Adipocyte differentiation of 3T3F442A cells was generated by treating confluent preadipocytes (day 0) with a differentiation medium (see "Experimental Procedures"). This treatment led to a  rapid (4 days after confluence) emergence of specific genetic markers of differentiation such as adipocyte fatty acid-binding protein (aP2) and hormone sensitive lipase (HSL) mRNAs (Fig.  4B), followed (7 days after confluence) by the accumulation of triglycerides (Fig. 4A), the final adipocyte phenotype. Whereas aP2 and HSL mRNA levels reached a plateau after 10 days (Fig. 4B), triglyceride content continued to increase regularly thereafter (Fig. 4A). Under these conditions, ATX mRNAs were detected 7 days after confluence, and their level increased regularly thereafter (Fig. 4B). Interestingly, the emergence of ATX mRNAs was delayed by 3-4 days as compared with that of aP2 and HSL mRNAs.
Up-regulation of ATX mRNA was tightly associated with the increase in LPLDact (Fig. 4C) and LPA concentration (Fig. 4D) in the culture medium. In parallel, the culture medium also contained LPC (the substrate for LPLDact), which was present at higher concentration than LPA but which was not significantly modified during adipocyte differentiation (Fig. 4D).
Up-regulation of ATX was also observed after adipocyte differentiation of the primary culture of preadipocytes. As described previously (24), the preadipocytes present in the stroma vascular fraction of adipose tissue can be differentiated into adipocytes when cultured for 7-10 days in the presence of a differentiating mixture (see "Experimental Procedures"). As shown in Fig. 5, after 10 days the expression of ATX mRNA was increased 10-fold as compared with the first day of culture. These results clearly show the existence of a regulation of ATX with adipocyte differentiation. ATX mRNA was also present in mature adipocytes isolated from mouse adipose tissue, and its expression was lower than in 10-day-old differentiated adipocytes (Fig. 5).
Recombinant ATX Allowed the Release of LPA by Preadipocytes-Our group has demonstrated previously that LPA is able to increase preadipocyte proliferation (5,6). Conversely to adipocytes, preadipocytes did not release LPLDact (see Fig. 4C) or LPA (Fig. 4D) but released LPC (see Fig. 4D), the substrate of the LPLDact of ATX. Therefore, we tested whether ATX could exert an LPA-dependent increase in preadipocyte proliferation.
To test this hypothesis, conditioned media from COS-7 cells transfected with empty pcDNA or pcDNA-mATX-FLAG were concentrated and used to treat 3T3F442A preadipocytes. As shown in Fig. 6A, 5 h of treatment with concentrated conditioned medium from pcDNA-mATX-FLAG transfected COS-7 cells (ATX-CCM) led to a dilution-dependent increase in LPA concentration in 3T3F442A preadipocyte culture medium. LPA was significantly detected with a 10 l/ml dilution and reached a maximum at 100 l/ml with a range of LPA concentration varying from 13 to 130 nM. Conversely no significant concentration LPA was generated after treatment with concentrated conditioned medium from empty pcDNA transfected COS-7 cells (sham CCM) (Fig. 6A).
To test the influence of recombinant ATX on preadipocyte proliferation, 3T3F442A preadipocytes were exposed to 50 l/ml sham CCM or ATX CCM, and the cell number was determined after 72 h of treatment. As shown in Fig. 6B, sham CCM led to an increase in cell number, suggesting the presence of a growth factor different from ATX in sham CCM. Nevertheless, treatment with ATX CCM led to an additional increase in cell number (Fig. 6B), which was completely abolished by cotreatment phospholipase B, a lysophospholipase that, as we and others have demonstrated previously, hydrolyzes and inactivates LPA (5,6,9,32). These results strongly supported the idea that recombinant ATX was able to increase preadipocyte proliferation according to an LPA-dependent mechanism.
Up-regulation of Adipocyte ATX in db/db Mice-To test the existence of a regulation of ATX expression with obesity, adipocyte ATX expression was measured in genetically obese diabetic mice (db/db). As expected, 12-week-old db/db mice were obese as attested by higher body weight (1.7-fold) and higher adipose tissue weight (9.5 and 10-fold in perigonadic and inguinal depot, respectively) when compared with their lean siblings, db/ϩ mice (Table II). As was also expected, db/db mice were diabetic as attested by their higher (3.6-fold) non-fasting glycemia when compared with db/ϩ mice (Table II). As shown in Fig. 7A, ATX mRNA expression was significantly higher in  5. Differentiation-dependent regulation of ATX in primary mouse preadipocytes. Preadipocytes were isolated from the stroma vascular fraction of the adipose tissue of a 4-week-old C57BL6J mouse as described under "Experimental Procedures." ATX mRNA expression was measured after overnight attachment (day 0) and after 10 days of culture (day 10) in a differentiating medium (see "Experimental Procedures"). ATX mRNA expression was also measured in mature adipocytes freshly isolated from mouse adipose tissue. Values are means Ϯ S.E. of three separate experiments. isolated adipocytes from db/db (3.8 and 4.8-fold in perigonadic and inguinal adipocytes, respectively) as compared with db/ϩ. A significant increase in ATX mRNA expression was also observed in interscapular brown adipose tissue (2.4-fold) and kidney (2.2-fold) (Fig. 7B). Conversely, no changes were observed in ATX mRNA expression in liver and brain (Fig. 7B). In a separate experiment, small explants of subcutaneous adipose tissue from db/ϩ and db/db mice were incubated in serum-free DMEM, and released LPLDact was measured. After 5 h of incubation, LPLDact released in the incubation medium was 18.2 Ϯ 2 and 29.3 Ϯ 4 pmol of LPA/mg of protein in db/ϩ (n ϭ 5) and db/db (n ϭ 5) adipose tissue explants (p Ͻ 0.01). This result showed that up-regulation of ATX mRNA in adipose tissue of db/db mice was accompanied by up-regulation of its activity. Taken together, the above results showed that db/db genetic obesity was associated with up-regulation of ATX expression in some tissues, including adipose tissue. DISCUSSION The initial objective of the present study was to identify the enzyme responsible for LPLDact released in the culture medium of adipocytes. Our results show the following: (i) that adipocyte LPLDact is co-purified with ATX; (ii) that ATX is expressed in adipocytes; (iii) that recombinant expression of ATX in COS-7 cells permits the release of LPLDact in their culture medium; (iv) that ATX expression and activity are strongly increased during adipocyte differentiation; (v) that recombinant ATX is able to increase preadipocyte proliferation via a LPA-dependent mechanism; and (vi) that adipocyte-ATX is substantially over-expressed in a genetic model of obesity.
ATX is an ecto-nucleotide pyrophosphatase phosphodiesterase initially discovered as a motility-stimulating protein in the culture medium of A2058 melanoma cells (25). ATX was then found to be expressed in several cell lines and tissues (28). The motility-stimulating action of ATX is blocked by pretreatment with pertussis toxin, suggesting the involvement of a G-protein-coupled receptor in its action (31). ATX is an ecto-protein anchored to the plasma membrane by a short N-terminal sequence and likely released in the extracellular medium after proteolytic cleavage. In its extracellular domain, ATX contains a large phosphodiesterase catalytic site involved in the hydrolysis of phosphodiester and pyrophosphate bonds present in a classical substrate such as ATP or ADP (26,30,33).
Our results strongly support the idea that ATX is responsible for the LPLDact released from 3T3F442A adipocytes. LPLDact was indeed co-purified with a protein exhibiting 100% homology with previously cloned mouse ATX. In addition, a release of LPLDact in culture medium was obtained after recombinant expression of mouse ATX cDNA in COS-7, a cell line that does not normally release LPLDact. Finally, ATX expression was clearly demonstrated in adipocytes and was tightly co-regulated with the release of LPLDact and LPA in culture medium during differentiation of 3T3F442A preadipocytes. Our conclusions are in agreement with recent publications that appeared during the completion of the present study, reporting that ATX was responsible for a LPLDact present in bovine and human sera (34,35).
ATX expression is strongly up-regulated during adipocyte differentiation. Terminal differentiation of adipocytes is a complex process that requires the emergence of specific genes such as aP2 and HSL (analyzed in the present work), which are necessary for the adipocytes to ensure its specific metabolic functions such as triglyceride accumulation. When comparing with aP2 and HSL, ATX expression is rather late during adipose conversion and is more closely associated with the accumulation of triglycerides. These observations suggest that ATX expression is a consequence of, rather than a requirement for, adipocyte differentiation. Further experiments will be neces-  6. Influence of recombinant ATX on preadipocytes. A, serum starved 3T3F442A preadipocytes were treated with decreasing dilutions of CCM from COS-7 cells transiently transfected with either empty pcDNA (sham-CCM) or pcDNA-mATX-FLAG (ATX-CCM) in DMEM supplemented with 1% BSA. After 5 h of treatment, LPA was quantified in preadipocyte culture medium using a radioenzymatic assay. B, serum-starved 3T3F442A preadipocytes were treated with 50 l/ml sham-CCM or ATX-CCM in DMEM supplemented with 1% BSA. Treatments were performed in the presence or absence of 0.1 units/ml phospholipase B (PLB). After 72 h of treatment, preadipocyte number was measured using a cell counter. sary to precisely identify the factor(s) involved in the regulation of ATX expression with adipocyte differentiation.
ATX expression in adipocytes is substantially increased in genetically obese diabetic db/db mice. These mice bear a mutation in the leptin receptor resulting in hyperphagia and the massive accumulation of adipose tissue associated with type II diabetes and, ultimately, type I diabetes. Up-regulation of adipocyte ATX in db/db suggests that this protein could either be associated with obesity or diabetes or both. ATX belongs to the ecto-nucleotide pyrophosphate phosphodiesterase (ENPP) family. This family encompasses the membrane glycoprotein plasma cell 1 (PC-1), a protein with a high structure homology with ATX (36). It is interesting to note that PC-1 expression was found to be increased in adipose tissue of obese diabetic Zucker fatty rats (37) which, like db/db mice, bear a mutation on the leptin receptor. Alterations in PC-1 expression have been associated with alterations in whole body insulin sensitivity as well as insulin receptor-tyrosine kinase activity (38,39). Further experiments will be necessary to identify the factor(s) involved in regulation of ATX in obesity and/or diabetes.
What could be the role of adipocyte-ATX? In adipose tissue, adipocytes and preadipocytes are present in the same environment and communicate with each other via paracrine mediators. As demonstrated in the present study, ATX catalyzes a LPLDact, which is responsible for the presence of LPA in adipocyte culture medium. We demonstrated previously that LPA is present in adipocyte culture medium and can activate preadipocyte growth via activation of the LPA1 receptor (6). In contrast to adipocytes, preadipocytes do not release ATX but release high concentration of LPC (see Fig. 4), the substrate of ATX for LPA synthesis. It can be speculated that, in intact adipose tissue, one of the possible roles of adipocyte ATX would be to exert a paracrine control on preadipocyte growth because of the generation of LPA at the extracellular face of the preadipocytes in close contact with adipocytes. This hypothesis is supported by our results showing that recombinant ATX is able to activate preadipocyte proliferation. In addition, the blocking effect of phospholipase B strongly suggests that the action of ATX on preadipocytes is very likely to be mediated by synthesis of LPA and the activation of the LPA receptor present at the surface of the preadipocytes. This hypothesis is in agreement with previous reports showing that ATX activates cell motility and cell proliferation in several cell lines via a G-proteincoupled receptor that could be a LPA-receptor (31,34).
What could be the consequence of the proliferative activity of ATX on preadipocytes? Preadipocyte proliferation is a key step in adipose tissue development, because it conditions the number of potential new adipocytes in the adipose tissue. In most obesity models such as db/db mice (where we observed an over-expression in adipocyte-ATX), the increase in adipose tissue mass results from an excess accumulation of triglycerides in existing adipocytes (hypertrophy), followed, above a certain level of hypertrophy, by the recruitment of new adipocytes (hyperplasia). It could be proposed that adipocyte ATX plays a role in the recruitment of new adipocytes by exerting a paracrine control on preadipocyte growth. ATX could therefore play a role in obesity-associated hyperplasia. This hypothesis is reinforced by very recent report (40) showing that LPA behaves as an agonist of the peroxisome proliferator activated receptor ␥, a crucial transcriptional factor involved in the adipocyte differentiation program.
Obesity-associated increase in fat mass is also dependent on its vasculature and therefore requires angiogenesis (41). ATX has previously been demonstrated to behave as a pro-angiogenic mediator (42). Adipocyte ATX could then play a role in the paracrine control of angiogenesis in adipose tissue. Finally, an endocrine action of adipocyte ATX can also be proposed. As pointed out earlier, ATX has recently been identified in blood (serum and plasma) (34,35), but, despite the presence of ATX mRNA in different tissues (28), the origin of circulating ATX remains unknown. When comparing it with other tissues, ATX mRNA appears to be fairly abundant in adipose tissue. In addition, considering the high mass of adipose tissue in the body, particularly in obese individuals, its significant contribution in circulating ATX can be proposed. Future development of transgenic animals exhibiting altered expression of adipocyte ATX will help to verify these different hypotheses.
In conclusion, the present work identifies ATX as a new adipocyte mediator exerting its biological activities via neosynthesis of the bioactive phospholipid LPA. Its close association with adipocyte differentiation and obesity and/or diabetes suggests a possible role of ATX in the normal or pathological development of adipose tissue.