Placental Thrombosis and Spontaneous Fetal Death in Mice Deficient in Ethanolamine Kinase 2*

Ethanolamine kinase catalyzes the first step in the CDP-ethanolamine pathway for the formation of the major membrane phospholipid phosphatidylethanolamine (PtdEtn). In this work, the predicted Etnk2 cDNA was established as a soluble protein with ethanolamine-specific kinase activity that was most highly expressed in liver. Mice with an inactivated Etnk2 gene were derived, and its absence reduced the rate of PtdEtn synthesis from exogenous ethanolamine in hepatocytes. PtdEtn is a major precursor to phosphatidylcholine in liver; however, Etnk2-/- mice did not have reduced amounts of either PtdEtn or phosphatidylcholine or an altered phospholipid molecular species distribution. The knock-out animals were able to adapt to a choline-deficient diet. The Etnk2-/- mice exhibited a maternal-specific intrauterine growth retardation phenotype that resulted in a 33% reduction in litter size and frequent perinatal death. Histological analysis of pregnant Etnk2-/- females showed that fetal development failed at the late stage of pregnancy in a significant percentage of embryos because of the appearance of extensive placental thrombosis. These results illustrate a non-redundant role for EtnK2 expression in regulating placental hemostasis.

There are two pathways for the synthesis of phosphatidylethanolamine (PtdEtn), 2 a major mammalian membrane phospholipid. The CDP-ethanolamine (Etn) pathway begins with ethanolamine kinase (EtnK; ATP:ethanolamine phosphotransferase, EC 2.7.1.82) followed by CTP:phosphoethanolamine cytidylyltransferase and terminates with a diacylglycerol phosphotransferase (1). The second route to PtdEtn is via the decarboxylation of phosphatidylserine (1,2). Although the CDP-Etn pathway is considered a major route for PtdEtn synthesis in most mammalian tissues (3), the phosphatidylserine decarboxylation pathway is the predominant route for PtdEtn in cultured cells (4 -7). The focus of our work is the first step in the CDP-Etn pathway, EtnK. There are four related mammalian genes that encode proteins possessing EtnK activity (see Fig. 1). Two of these, choline kinase (ChoK) ␣ and ChoK␤ (see Fig. 1), are characterized as ChoKs, but also are able to phosphorylate Etn, a choline (Cho) analog, to a lesser extent (8). Based on this property, both the ChoK and EtnK reactions were considered for some time to be carried out by the same enzyme(s). However, in Saccharomyces cerevisiae, there are two kinases with opposite specificities. The yeast ChoK has a specific activity 3.6-fold higher for Cho compared with Etn, whereas EtnK has a specificity constant 2-fold higher for Etn compared with Cho (9). The discovery of the Drosophila EtnK encoded by the eas gene established the existence of Etnspecific kinases with negligible ChoK activity in higher eukaryotes (10). A defect in the Drosophila EtnK results in a bang-sensitive paralytic phenotype, indicating the potential importance of EtnK to neuronal function (10), and EtnK is also implicated in fly development (11). Mice and humans have two related EtnK genes (see Fig. 1). Human ETNK1 has robust EtnK activity, but ChoK activity is not detected (12). Expression of ETNK1 in cell culture selectively increases Etn uptake and phosphorylation and PtdEtn synthesis (12). A second gene (ETNK2) is identified as a cDNA related to ETNK1 (12). Although bioinformatics analysis suggests that ETNK2 encodes EtnK-specific kinase activity, there are no biochemical data to support this conclusion. ETNK2 is the most highly expressed of the two isoforms and is found primarily in the liver and reproductive tissues of humans (12). A role for EtnK2 in testicular maturation is suggested by the pattern of Etnk2 expression in developing mouse testis (13). Thus, EtnKs can be distinguished from ChoKs, and a number of possible roles from the literature are suggested for mammalian EtnKs.
This investigation establishes mouse EtnK2 as a specific EtnK that is expressed most highly in liver and reproductive tissues. Characterization of phospholipid synthesis in the absence of EtnK2 expression in mice pointed out its dispensability or redundancy in most physiologic processes and cell types. Homozygous disruption of the Etnk2 gene had little effect on liver phospholipid metabolism, neural development, or testicular function. However, a non-redundant role for EtnK2 was indicated by a maternal-specific failure to support late embryonic development, resulting in reduced perinatal size and survival and suggesting compromised placental function. This phenotype was similar to that of the ␤ 2 -glycoprotein I null mouse, which lacks the principal target antigen for anti-phospholipid antibodies associated with recurrent pregnancy loss and fetal growth restriction in women (14). It is thought that the ␤ 2 -glycoprotein I protein complexes with anionic phospholipids and that these complexes can become antigenic and underlie a novel class of autoimmune disorders called antiphospholipid syndrome. Anti-PtdEtn antibodies in particular are prevalent in antiphospholipid syndrome patients with mid-tolate pregnancy loss (15). The Etnk2 ؊/؊ phenotype also resembles the Rho-associated kinase II (ROCK-II) null mouse, which lacks expression of the Rho-associated kinase II isoform, and both male and female embryos exhibit extensive thrombus formation in the placenta, leading to intrauterine growth retardation and low birth weight pups (16). ROCK-II is a serine/threonine kinase that phosphorylates cytoskeletal proteins upon binding to the active form of the Rho GTPase (17). RhoA cycling from the active GTP form to the inactive GDP form can be regulated by the lateral movement of PtdEtn at the cell surface during late stage cytokinesis (18,19). These new data in this work characterize mammalian EtnK2 for the first time and suggest a link between maternally derived PtdEtn and embryonic signaling across the placental interface.

EXPERIMENTAL PROCEDURES
Cloning of the Mouse Etnk2 cDNA and Construction of the Expression Vector-The expressed sequence tag data base was searched using the human ETNK1 sequence (12). Two mouse expressed sequence-tagged clones (AI098609 and AA261067) that encoded amino acid 25 through the termination stop codon were identified. The 554-bp EcoRI-NsiI fragment of AI098609 was ligated to the 4352-bp EcoRI-NsiI sequence of AA261067 in the pT7T3D-Pac vector, and these fragments together made up the 3Ј-end of the cDNA. The 5Ј-end of the cDNA was amplified from a mouse kidney MarathonReady cDNA kit (Clontech) using forward primer 5Ј-ACCACTCC-CACTCTCTCTCAGG and reverse primer 5Ј-CATAC-ACCCGGACCAACACG and cloned into the pCR2.1 vector. The 5Ј-end was sequenced and then subcloned into pcDNA3.1 after XbaI-BamHI digestion. The 3Ј-end was ligated with the 5Ј-end in the expression vector at the AfeI-HindIII sites, resulting in plasmid pPJ184, which carried the full-length mouse Etnk2 cDNA. The entire cDNA sequence was confirmed in both directions.
Transcript Analysis-A mouse multiple tissue Northern blot (OriGene) loaded with equal amounts of tissue poly(A) ϩ mRNA was hybridized and washed according to the manufacturer's instructions. The blot was hybridized with 32 P-labeled random-primed probes prepared using the homologous 512-bp NsiI-ScaI fragment of the mouse Etnk2 cDNA (12) as template. Probes were made using random primers and [ 32 P]dCTP.
Quantitative Real-time PCR-The abundance of Etnk2 mRNA was measured by real-time PCR as described previously (20). The oligonucleotide primers and probes were synthesized at the Hartwell Center for Bioinformatics and Biotechnology at St. Jude Children's Research Hospital. The primers and probes were used to assess expression of Etnk2 mRNA (see Table 1) and the glyceraldehyde-3-phosphate dehydrogenase housekeeping gene (control reagent kit, Applied Biosystems). Amplification and detection of the real-time PCR products were performed with TaqMan Universal PCR Master Mixes (Applied Biosystems) and an ABI PRISM 7700 sequence system (Applied Biosystems). Experiments were performed with three replicas for each cDNA, and negative reverse transcriptase and no-template controls were run for each reaction on the same 96-well plate. All of the real-time values were compared using the C T method, in which the amount of target mRNA was normalized to the glyceraldehyde-3-phosphate dehydrogenase housekeeping gene (⌬C T ).
Cell Culture-Human kidney 293T cells (a gift from Suzanne Baker, Department of Developmental Neurobiology, St. Jude Children's Research Hospital) were cultured in Dulbecco's modified Eagle's medium (BioWhittaker) supplemented with 10% fetal bovine serum (Atlanta Biologicals, Inc.), 2 mM glutamine, and 10 mM HEPES (pH 7.0) in a 5% CO 2 humidified atmosphere at 37°C. Transient transfection of pPJ184 expressing EtnK2 protein, pAL10 expressing human ETNK1, or the pcDNA3.1 empty vector control was performed using FuGENE 6 (Roche Applied Science) according to the manufacturer's instructions. Cells were scraped from the dishes 48 h after transfection after washing with phosphate-buffered saline (PBS) and centrifuged, and the supernatant was removed. Cell pellets were stored frozen at Ϫ80°C.
Construction of the Mouse Etnk2 Replacement Vector-In general, the Neo resistance gene cassette was inserted into exon 3 of the gene to replace 66 bp. A negative selection cassette encoding the diphtheria toxin antigen was added at the 3Ј-end of the mouse genomic DNA insert. Using the mouse Etnk2 cDNA sequence, the genomic Etnk2 sequence (AC034108) was identified in the NCBI mouse genomic data base. The short and long arms of the construct were obtained by PCR using the Advantage genomic PCR kit (Clontech) following the manufacturer's directions. Mouse W9.5 embryonic stem (ES) cell genomic DNA was the template; oligonucleotide primers 5Ј-TGTCCGCGGCACCAGTCAACCCACATAAGTC (including a SacII site) and 5Ј-GCATCTAGAGTGGATGGTGTGAATC-TTAGCC (including a XbaI site) yielded a 1903-bp product, and 5Ј-CGTATCGATGATGAGATCAGTCCCAGGTATA-GAGA (with a ClaI site) and 5Ј-TGGGTCGACTGGTTGCC-AATGTCAAATGCCT (with a SalI site) yielded a 5752-bp product. PCR products were purified using a High Pure PCR product purification kit (Roche Applied Science) and then incubated at 72°C for 20 min after the addition of dATP and Taq polymerase (Promega). This ensured the presence of an A overhang for the subsequent cloning into the pCR2.1 vector. The sequences of the long and short arms were confirmed for 650 bases at each end. The plasmid pMAK6 (pBluescript SK ϩ containing the diphtheria toxin antigen cassette; a gift from Dr. James Ihle, Department of Biochemistry, St. Jude Children's Research Hospital) was used as the base vector. A 1.2-kb Neo cassette (a gift from Dr. James Ihle) was inserted into the ClaI and XbaI sites, followed by the 1889-bp SacII-XbaI short arm fragment and the 5739-bp ClaI-SalI long arm fragment. The final plasmid construct was linearized with SacI prior to transfection into mouse AB2. 2

ES cells (Lexicon Genetics).
Identification of Gene-targeted ES Cell Clones-ES cells were grown on mitotically inactivated mouse embryonic fibroblasts carrying resistance to neomycin (a gift from Peter McKinnon, Department of Genetics, St. Jude Children's Research Hospi-tal). More than 200 clones resistant to G418 were selected; genomic DNA was isolated and screened by PCR using primers FP1, RP1, and RN1 (see Table 1). ES cells with one chromosome that had undergone homologous recombination with the replacement vector were identified by a 3366-bp PCR product compared with a 2261-bp product from the wild-type allele obtained using primers FP1 and RP1. Insertion of the Neo cassette was confirmed by a 2479-bp product obtained with primers FP1 and RN1.
Derivation of Etnk2-disrupted Mice-An ES cell colony containing the recombined Etnk2-disrupted locus was injected into C57BL/6J blastocysts, which were then implanted into pseudopregnant female mice at the Transgenic Core Facility of St. Jude Children's Research Hospital. Male offspring with 75-90% agouti color (the coat color contributed by the ES cells) were bred with C57BL/6J females. Pups that were 100% agouti, indicating germ line transmission, were screened. Tail clips from weanling animals were digested overnight at 55°C in buffer containing 0.25 mg/ml proteinase K, 0.2 M NaCl, 100 mM Tris-HCl (pH 8.5), 5 mM EDTA, and 0.2% SDS. The DNA was extracted using an equal volume of phenol/chloroform/isoamyl alcohol, and the upper phase was precipitated with 0.8 volume of cold isopropyl alcohol. The DNA precipitate was washed with 80% ethanol and dissolved in 50 -100 l of water at 50°C for 30 min. The DNA was then subjected to multiplex PCR analysis using primers FS1, RS1, and RN3 (Table 1), which yielded a 325-bp fragment for the recombinant allele and a 225-bp fragment for the wild-type allele. Etnk2 ϩ/Ϫ heterozygotes were mated, and the progeny were screened by the same methodology. Homozygous mouse pups were identified at an ϳ25% ratio and bred together to maintain the colony. Mice were maintained on regular Purina rodent chow 5013 at a room temperature of 72 Ϯ 1°F and a room humidity of 50 Ϯ 10% with a 12-h light/12-h dark cycle, with the dark cycle starting at 6:00 p.m.
Tissue Histology-Mouse tissues were analyzed using both paraffin sections and frozen sections. Mice were anesthetized with isoflurane and intracardially perfused with 0.01 M PBS followed by 4% paraformaldehyde. Tissues (including liver, ovary, uterus, brain, retina, and cochlea) were dissected out and incubated overnight in 4% paraformaldehyde at 4°C. For frozen sections, tissues were immersed in 30% sucrose for 24 -48 h at 4°C and frozen on dry ice in embedding medium (Fisher). Sixmicrometer-thick sections were cut and stored at Ϫ20°C until used. For paraffin sections, mouse tissues were fixed in 10% formalin and processed by dehydration in 70% ethanol followed by xylene. Tissues were infiltrated and embedded in paraffin, cut into 4-m sections, mounted on microscope slides, and stored at room temperature until used. Sections were stained with hematoxylin and eosin and viewed under an Olympus BX41 microscope.
Six-micrometer-thick cryosections were used for programmed cell death analysis. Sections were stained based on reactivity determined by terminal deoxynucleotidyltransferase-mediated dUTP nick end labeling (TUNEL) using the ApopTag peroxidase in situ apoptosis detection kit (catalog no. S7100, Chemicon) according to the manufacturer's instructions. In addition, active caspase-3 activity was detected using an antibody that specifically interacts with the cleaved activated fragment of caspase-3. Immunohistochemistry was performed after permeabilization of the cells with 0.2% Triton and blocking in 2.5% horse serum in PBS. Sections were incubated overnight with a 1:200 dilution of a polyclonal antibody against cleaved caspase-3 (Abcam) at 4°C. The next day, sections were washed with PBS and incubated at room temperature with an Alexa 488-conjugated secondary antibody (Molecular Probes) for 1 h. After washing and mounting, the sections were viewed under an Olympus BX41 fluorescence microscope.
Hormone Determinations-Serum samples were stored at Ϫ20°C until analysis. Coat-A-Count radioimmunoassays (Diagnostic Products Corp.) were performed to quantitatively measure the amounts of progesterone and estradiol in serum samples according to the manufacturer's instructions. Briefly, 100 l of duplicate serum samples or known concentrations of hormone standards were added to tubes coated with antibody. 125 I-Labeled progesterone or estradiol was then added to each tubes and vortexed. Samples were incubated at room temperature for 3 h, decanted, and air-dried. Tubes were counted for 1 min in a Beckman ␥-counter.
EtnK and ChoK Assays-Frozen 293T cell pellets transfected with plasmid pPJ184 or pPJ96 (12) or with a control empty vector were thawed on ice and incubated for 1 h in lysis buffer (20 mM Tris-HCl (pH 8.0), 2 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml leupeptin, and 2 g/ml aprotinin) on ice. The cells were disrupted by sonication (6 ϫ 30 s) in a cup horn (Microsonix Sonicator 3000), and the particulate fraction was removed by centrifugation at 5000 ϫ g for 5 min at 4°C. Supernatants were transferred to new tubes, and EtnK or ChoK activities were determined essentially as described previously (12,21). A flash-frozen tissue slice was thawed on ice, and the wet weight was determined. Two volumes of lysis buffer were added to ϳ50 mg, and the tissue was broken in a glass homogenizer by 20 strokes and set on ice for 30 min, followed by 15 more strokes. The lysate was centrifuged at 5000 ϫ g for 10 min at 4°C, and the supernatant was transferred to a new tube prior to assay. The standard assays contained 100 mM Tris-HCl (pH 8.0), 10 Knock-out Mice reaction mixtures were incubated at 37°C for 15 min when using transfected cells or for 30 min when using the mouse tissue lysates. The reaction was stopped by the addition of 10 l of 0.5 M Na 3 EDTA, and the tubes were vortexed and placed on ice. Next, 40 l of each sample were spotted on pre-adsorbent Silica Gel G thin-layer plates, which were developed with 2% ammonium hydroxide and 95% ethanol (5:1, v/v). Phosphoethanolamine and phosphocholine were identified by co-migration with standards, and product formation was quantitated using a Bioscan imaging system. Protein was determined according to the Bradford method (22). Phospholipid Determinations-Flash-frozen mouse tissues were thawed and weighed, and ϳ50 mg were extracted by the method of Bligh and Dyer (23). The organic phase containing lipid was concentrated under nitrogen and resuspended in 400 l of chloroform/methanol (2:1). A 1-l aliquot was loaded onto a thin-layer silica gel rod, developed first in ether, dried, and then developed in chloroform/methanol/acetic acid/water (50:25:8:3). Lipid mass was detected by flame ionization using an Iatroscan instrument (Iatron Laboratories, Inc., Tokyo, Japan) with PeakSimple software (SRI Instruments), and peaks were identified by co-migration with authentic standards. PtdEtn, phosphatidylcholine (PtdCho), cholesterol, or triglyceride mass was calculated by consulting standard curves for detector response prepared with dioleoyl-PtdEtn, egg PtdCho, cholesterol, or triolein (Matreya, LLC), respectively.
Phospholipid Molecular Species Profiling by Mass Spectrometry-Mass spectrometry analysis was performed in the Hartwell Center for Bioinformatics and Biotechnology at St. Jude Children's Research Hospital using a Finnigan TM TSQ Quantum triple quadrupole mass spectrometer (Thermo Electron Corp., San Jose, CA) equipped with a nanospray ion source. Samples were introduced in 50:50 (v/v) chloroform/ methanol and 1% formic acid via static nanoelectrospray using EconoTips TM (New Objective, Inc., Woburn, MA). The instrument was operated in the positive ion mode using parent ion scanning for PtdCho and neutral loss scanning for PtdEtn corresponding to the loss of their respective phosphoryl head groups. Ion source parameters were as follows: spray voltage, 1000 V; capillary temperature, 270°C; and capillary offset, 35 V. Tube lens offset was set by infusion of the polytyrosine tuning and calibration solution (Thermo Electron Corp.) in the electrospray mode. Acquisition parameters for PtdCho were as follows: scan range, m/z 600 -900; scan time, 0.3 s; product mass, m/z 184.1; collision energy, 40 V; peak width Q1 and Q3, 0.7 full-width half-maximum; and Q2 collision-induced dissociation gas, 0.5 millitorr. Acquisition parameters for PtdEtn were as follows: scan range, m/z 600 -900; scan time, 0.3 s, neutral loss mass, m/z 141.0; collision energy, 30 V; peak width Q1 and Q3, 0.7 full-width half-maximum; and Q2 collision-induced dissociation gas, 0.5 millitorr. Instrument control and data acquisition were performed with Finnigan TM Xcalibur TM software (Version 1.4 SR1, Thermo Electron Corp.).
Preparation and Radiolabeling of Hepatocytes-Whole livers of 2-4-month-old mice were harvested and rinsed in ice-cold buffer A (each liter contained 3.9 g of NaCl, 0.5 g of KCl, 24 g of HEPES, and 2.7 g of glucose, with the pH adjusted to 7.6 with NaOH). Each liver was chopped with a Vibratome tissue chopper at a setting of 0.5 mm and resuspended in 20 ml of erythrocyte buffer (15 ml of buffer A plus 5 ml of erythrocyte lysis buffer (Qiagen Inc.)) in a 125-ml Erlenmeyer flask, gently mixed in the buffer, and allowed to settle. The supernatant was carefully poured off and discarded, and the procedure with the erythrocyte buffer was repeated twice. The tissue pieces were then resuspended in 30 ml of digestion buffer (buffer A plus 0.7 g/liter CaCl 2 , 0.5 mg/ml type I collagenase, and 6 g/ml deoxyribonuclease) and incubated in a water bath shaker at 120 rpm for 20 min at 37°C. The tissue pieces were allowed to settle on ice, and the supernatant was transferred to a centrifuge tube. The tissue pieces were resuspended in another 30 ml of digestion buffer; the procedures were repeated twice; and the three supernatants were combined. The cell suspension from the supernatants was filtered through Spectra/Mesh nylon (41 m; Fisher), and liver cells in the filtrate were centrifuged down at 100 ϫ g for 2 min at room temperature. The cell pellet was resuspended in wash buffer (each liter contained 8 g of NaCl, 0.35 g of KCl, 0.16 g of MgSO 4 , 0.18 g of CaCl 2 , 2.4 g of HEPES, and 15 g of bovine serum albumin, with the pH adjusted to 7.4 with NaOH) containing 10% (v/v) erythrocyte lysis buffer and centrifuged down three times.
Mouse liver cells were resuspended in PBS containing 0.5% bovine serum albumin to a density of ϳ1 ϫ 10 7 cells/ml. For each labeling experiment, 75 l of the cell suspension were used. The cells were labeled with 2 M [ 3 H]Etn or [ 3 H]serine (50 Ci/mmol). After a 6-h incubation at 37°C, cells were harvested and washed twice with PBS. The cells were lysed by sonication in 50 l of buffer containing 20 mM Tris-HCl (pH 7.5), 2 mM dithiothreitol, 5 mM EDTA, and 50 mM NaF. Cell debris was removed by centrifugation, and aliquots were removed for determination of protein and total [ 3 H]pantothenate incorporated into the cell. The identity of the intracellular metabolites was determined by fractionating the cell lysates by thin-layer chromatography on Silica Gel G plates developed with 2% ammonium hydroxide and 95% ethanol (5:1) using the standards described previously (24). Sections (0.5 cm) were scraped from the plate, and the radioactivity was quantitated in 3 ml of scintillation fluid using a liquid scintillation counter.

RESULTS
Sequence and Distribution of the Murine Etnk2 cDNA-In our previous work with ETNK1, a second human gene called ETNK2 was detected by a bioinformatic analysis (12). The mouse Etnk2 gene is located on chromosome 1 and consists of eight exons that span 16.7 kb. The full-length cDNA was assembled as described under "Experimental Procedures," and the complete 5Ј-end is contained in expressed sequence tag XM_903670 (reference sequence NT_039192). There is an in-frame stop codon upstream from the predicted ATG initiation codon. The version of ETNK2 used throughout this study was isoform 3 (XM_903670) in the NCBI Database. It is important to note that the mouse Etnk2 reference sequence (isoform 1, NP_780652.1) in the NCBI Database does not correspond to the protein we are working with. This sequence lacks 28 amino acid residues at the N terminus and lacks the C-terminal domain that is conserved in all ChoKs and EtnKs. Specifically, the reference sequence lacks the essential aspartate Etnk2 Knock-out Mice SEPTEMBER 22, 2006 • VOLUME 281 • NUMBER 38

JOURNAL OF BIOLOGICAL CHEMISTRY 28441
residue that coordinates the Mg 2ϩ in ATP binding and the conserved structural motif (ILV)X 2 ID(FWY)E(YF)X 3 NX 3 (FYW)DX-6 E, which is critical for catalysis in this protein family (25). Thus, the isoform 1 transcript will not produce catalytically active EtnK. Isozyme 2 (AAP47267) has a truncated N terminus that begins between isoforms 1 and 4, and isozyme 4 (XM_903847) is even shorter, lacking the first 103 residues at the N terminus of isozyme 3. It is not known if these significant deletions at the N terminus, which remove a significant number of residues conserved in EtnK1 (Fig. 1A), would encode a functional EtnK2. EtnK2 isoform 5 (XM_903848) is a splice variant of isozyme 3 with exon 7 and the conserved ChoK/EtnK signature motif deleted and is strongly predicted to lack activity. We selected isoform 3 as the reference EtnK2 because it has the longest open reading frame.
Northern blot analysis showed that Etnk2 was most highly expressed in liver, followed by kidney and testis (Fig. 1B). The Northern blot signals a transcript of 2.3 kb, which is consistent with the size of our cloned cDNA (Fig. 1A). The distribution of expressed sequence tags is consistent with our Northern analysis and shows that the ovary expresses about the same level of Etnk2 as the testis and that the eye and brain have equivalent expression levels (NCBI EST Profile Viewer). The tissue distribution of mouse Etnk2 mRNA (Fig. 1B) mirrored the distribution of human ETNK2 transcripts (12). A quantitative view of the Etnk2 expression levels in various tissues was obtained by quantitative real-time PCR (Fig. 1C). Kidney Etnk2 transcript levels were 100-fold lower than liver transcript levels, and most of the other tissues were between 1000-and 100,000-fold lower compared with liver.
Catalytic Activity of EtnK2-The enzymatic activity of the Etnk2 cDNA clone was analyzed by transient expression in 293T cells (Fig.  2). The lysates from cells transfected with Etnk2 cDNA exhibited significantly higher levels of EtnK activity, but did not possess ChoK activity above the background levels detected in control cell lysates. Increasing the protein concentration in the assay by 10-fold did not reveal any evidence of increased ChoK activity (data not shown), placing the ChoK activity at Ͻ0.1% of the EtnK activity in EtnK2. The kinetic constants for EtnK2 were estimated in the lysates. The enzyme had an apparent high affinity for both Etn (37 M) and ATP (26 M). Thus, the Etnk2 cDNA encoded a robust EtnK activity that lacked detectable ChoK activity.  (12) are specific for Etn and do not use Cho as substrate. ChoK␣ and ChoK␤ selectively use Cho, but also phosphorylate Etn to a lesser extent (8). Residues in the other proteins that are identical to those in EtnK2 are highlighted. Alignment was performed using the program T-coffee (42). B, a multiple tissue Northern blot (OriGene) loaded with equal amounts of poly(A) ϩ mRNA was hybridized with a 32 P-labeled probe prepared from the 512-bp NsiI-ScaI fragment of the mouse Etnk2 cDNA. Sm. Intestine, small intestine. C, Etnk2 mRNA abundance in various tissues was analyzed by quantitative real-time PCR. All of the real-time values were compared using the C T method, in which the amount of target mRNA was normalized to the glyceraldehyde-3-phosphate dehydrogenase housekeeping gene (⌬C T ). The abundance in mouse liver was set to 1 as the calibrator, and the relative abundance in the other tissues was calculated.

Etnk2 Knock-out Mice
Generation of Etnk2 Ϫ/Ϫ Mice-The strategy for the inactivation of the Etnk2 allele is diagramed in Fig. 3. Gene function was eliminated by the introduction of the Neo selection cassette into the third exon of the gene. The Neo insertion into exon 3 eliminated residues 186 -207 of the protein sequence, removing a region corresponding to helix E in the ChoK/EtnK structure (25). The truncated protein lacked the ATP-binding site (Brenner's motif) plus the ChoK/EtnK signature motif and is catalytically inactive. The targeting construct also contained the diphtheria toxin antigen cassette used for negative selection in the isolation of ES clones (Fig. 3). The structure of the knock-out allele was verified by PCR. PCR screening was also the routine method employed to genotype the mice by the DNA extracted from tail clippings (Fig. 3, inset). The ES cells were selected and screened, and germ line transmission of the knock-out allele was accomplished as described under "Experimental Procedures." A line of homozygous Etnk2 Ϫ/Ϫ mice were derived from a breeding program.
EtnK Activity in Knock-out Mice-Etnk2 was most highly expressed in liver (Fig. 1B); therefore, our focus with the homozygous knock-out mice was to analyze the effect of ablation of Etnk2 on kinase activities and lipid metabolism in liver. Livers were removed from wild-type and Etnk2 Ϫ/Ϫ mice; soluble lysates were prepared; and the EtnK and ChoK activities were determined (Fig. 4). The liver lysates from the Etnk2 Ϫ/Ϫ mice possessed 10-fold lower specific activities for EtnK compared with those from the wild-type mice. We were always able to detect EtnK activity in the liver lysates from the knock-out mice. This residual EtnK activity was attributed to a combination of the contributions of liver ChoKs, which have some activity with Etn (8), and the EtnK1 enzyme. The ChoK activities were equivalent in the two groups of mice. We next analyzed a series of murine tissues for the presence of EtnK in the knock-out mouse model (Table 2). Clearly, the liver was the most affected of the tissues. Organs such as spleen, heart, and brain possessed the same levels of tissue EtnK activity in the Etnk2 Ϫ/Ϫ mice as in the wild-type mice, consistent with the low expression of Etnk2 in these tissues (Fig. 1B). One surprise was the modest reduction in testis EtnK activity because Etnk2 was reportedly highly expressed in this tissue (13). ChoK activities were not appreciably different in any of the tissues derived from the Etnk2 Ϫ/Ϫ mice compared with the wild-type animals ( Table 2). These enzymatic and expression data show that the liver experienced the most severe reduction in EtnK activity in the knock-out mouse model.
Phospholipid Composition of the Livers of Etnk2 Ϫ/Ϫ Mice-If EtnK2 is important for the synthesis of PtdEtn in liver, one possible result  Targeting construct for generating Etnk2 knock-out mice. A portion of the wild-type Etnk2 gene is diagramed (not to scale) above the targeting construct that contained the Neo gene inserted into exon (Ex) 3 in place of 63 bp of exon 3 coding sequence to function as the positive selection marker. The diphtheria toxin antigen cassette was inserted to the right of the long arm and functioned as the negative selection in the isolation of ES cells. The structure of the mutated allele illustrates the position of the primer pairs that were used in genotyping. Inset, PCR genotyping of Etnk2 knock-out mice. DNA was extracted from tail clippings and subjected to multiplex PCR using primers RS1, RN3, and FS1. The wildtype allele produced a single band of 225 bp; the knock-out mice produced a single band of 325 bp; and the heterozygous mice exhibited both bands. SEPTEMBER 22, 2006 • VOLUME 281 • NUMBER 38 would be a deficiency in this phospholipid or an alteration in its composition. We measured the amount of the two major liver phospholipids (PtdCho and PtdEtn) in the wild-type and Etnk2 Ϫ/Ϫ mice and found no difference in the tissue content of these two species (Table 3). We also examined the molecular species fingerprint of liver PtdEtn and PtdCho in the knock-out mice and found that there was no difference in the molecular species composition of PtdEtn or PtdCho between the wildtype and Etnk2 Ϫ/Ϫ mice (Fig. 5). The unique fatty acid composition of each phospholipid class observed in the wild-type mice was maintained in the knock-out animals. These data demonstrate that the absence of EtnK2 activity did not have a demonstrable effect on the amount or composition of the major liver phospholipids.

Etnk2 Knock-out Mice
PtdEtn synthesis is essential for the maintenance of liver function in mice on a Cho-deficient diet (26,27). Mice lacking the ability to convert PtdEtn to PtdCho via PtdEtn N-methyltransferase succumb to liver failure within 3-4 days of being switched to a Cho-deficient diet. We reasoned that PtdEtn production by EtnK2 through the CDP-Etn pathway may be critical for supplying the PtdEtn for methylation during Cho deprivation. A group of Etnk2 Ϫ/Ϫ and wild-type littermate control mice were switched to a Cho-deficient diet at 3 weeks of age and maintained on the diet for 8 weeks. Both groups of animals remained outwardly healthy during this period. The animals were killed, and the livers were examined. There was no difference in the liver histology in the knock-out and control animals (data not shown), and the hepatic content of PtdCho and PtdEtn was the same (Table 3). Thus, it was clear that sufficient PtdEtn was being produced for PtdCho formation during Cho deprivation in the Etnk2 Ϫ/Ϫ mice.
Phospholipid Synthesis in Etnk2-deficient Liver-The next step in our analysis was to determine whether the 10-fold reduction in EtnK activity in hepatocytes from Etnk2 Ϫ/Ϫ mice resulted in a deficiency in PtdEtn synthesis from Etn. The phosphorylation of Etn is the first step in the conversion of extracellular Etn to PtdEtn, and it was clear from our experiments that Etnk2 Ϫ/Ϫ hepatocytes were defective in PtdEtn synthesis (Fig.  6A). However, the rate of PtdEtn synthesis in Etnk2 Ϫ/Ϫ hepatocytes was only half the rate observed in the wild-type cells, illustrating that the 10-fold reduction in EtnK activity did not translate into an equivalent reduction in PtdEtn formation by the CDP-Etn pathway. We also performed a labeling experiment with [ 3 H]serine to determine whether there was a difference in the wild-type and knock-out hepatocytes in PtdEtn synthesis from phosphatidylserine (Fig. 6B). These data did not reveal a difference in PtdEtn synthesis from serine in the two hepatocyte preparations.
Reproductive Phenotype in Etnk2 Ϫ/Ϫ Mice-Although there was no discernable phenotype in the adult mutant mice, they did have a clear reproductive defect. Litters from crosses between Etnk2 Ϫ/Ϫ parents resulted in smaller litter sizes and pup mortality (Table 4). Matings between wild-type parents produced eight weaned pups/litter; however matings between Etnk2 Ϫ/Ϫ mice produced 33.6% fewer pups/litter ( p Ͻ 0.0001). This defective reproductive phenotype was maternal-specific. Matings of Etnk2 Ϫ/Ϫ males with wild-type females resulted in normal litter sizes of Etnk2 ϩ/Ϫ offspring (Table 4). In contrast, matings of Etnk2 Ϫ/Ϫ females with wild-type males resulted in low numbers of Etnk2 ϩ/Ϫ offspring ( p Ͻ 0.0001). A high rate of pup mortality contributed to the low number of pups produced.

. EtnK and ChoK activities in the livers of wild-type and
Etnk2 ؊/؊ mice. The livers of three wild-type and Etnk2 Ϫ/Ϫ mice were removed, and soluble lysates were prepared and assayed for EtnK and ChoK activities as a function of protein using the biochemical assays described under "Experimental Procedures." The mice were selected from 3-month-old littermates from a cross between heterozygous parents. EtnK activity in wild-type (F) and knock-out (E) liver extracts was compared with ChoK activity in wild-type (f) and knock-out (Ⅺ) liver extracts. The error bars are S.D. of triplicate determinations.

TABLE 2 EtnK and ChoK activities in mouse tissues
Mouse tissues were harvested from 3-month-old male wild-type or Etnk Ϫ/Ϫ mice and homogenized, and soluble lysates were prepared and assayed for EtnK and ChoK activities as described under "Experimental Procedures." There were three mice in each group, and the error ranges are S.D.

TABLE 3 PtdCho and PtdEtn content of livers from mice maintained on a normal chow diet and a Cho-deficient diet
The livers were obtained from 3-month-old littermates from a cross between two heterozygous parents. The animals on the Cho-deficient diet were placed on the diet at 3 months of age and maintained for 8 weeks prior to liver analysis. There were three mice in each group, and the error ranges are S.D.

Etnk2 Knock-out Mice
There were no pup deaths in the matings between wild-type mice or between wild-type females and Etnk2 Ϫ/Ϫ males (Table  4). In contrast, matings between wild-type males and Etnk2 Ϫ/Ϫ females ( p Ͻ 0.0001) and between Etnk2 Ϫ/Ϫ males and Etnk2 Ϫ/Ϫ females ( p Ͻ 0.0001) resulted in 10.5 and 16.9% pup mortality, respectively. Matings of Etnk2 ϩ/Ϫ females with Etnk2 ϩ/Ϫ males resulted in normal litter sizes; no pup mortality; and the expected distribution of wild-type, heterozygous, and knock-out animals. The disrupted Etnk2 allele was backcrossed onto the C57BL6/J background for eight generations, and the backcrossed homozygous null females had an average litter size of 5.4 (n ϭ 7), thus retaining the high rate of perinatal mortality. In summary, EtnK2 clearly plays a non-redundant role in the maternal nurturing of the embryo. Histology of Etnk2 Ϫ/Ϫ Mice-Tissue sections of liver, brain, retina, cochlea, ovary, and uterus from three individual Etnk2 Ϫ/Ϫ and wild-type mice at postnatal day 60 were stained with hematoxylin and eosin. The hepatocytes in liver lobes were well organized, and a similar morphology was also observed in Etnk2 Ϫ/Ϫ mice. The cellular structure of the brains from the Etnk2 Ϫ/Ϫ mice, including the internal granular layer, external granular layer, and Purkinje cell layer, appeared similar to those from the wild-type controls, illustrating that disruption of the cell layers in the cerebral cortex and hippocampus did not occur in Etnk2 Ϫ/Ϫ mice (data not shown). We also examined the cochlea and retina structures and observed no differences between Etnk2 Ϫ/Ϫ and control mice (data not shown). The mice were also tested for sensorimotor coordination in the accelerating rotarod test, which is sensitive to cerebellar and basal ganglial dysfunction. Etnk2 Ϫ/Ϫ mice of both genders showed equal latency to fall as the wild-type controls, suggesting a normal cerebellum (data not shown).
The female reproductive phenotype in Etnk2 Ϫ/Ϫ mice could not be attributed to altered morphology or hormone levels. Control mice had several developing follicles, multiple corpora lutea, and a well organized germinal epithelium; interstitial stromal cells and follicles were at various stages of development. The ovaries and uteruses from Etnk2 Ϫ/Ϫ mice, dissected at different stages of gestation, were the same as those from the wild-type controls. The progesterone levels in the pregnant Etnk2 Ϫ/Ϫ females (32.7 Ϯ 14 ng/ml serum, n ϭ 11) were not different from those in the pregnant wild-type females (31.5 Ϯ 12 ng/ml serum, n ϭ 9). Likewise, there was no difference in the estradiol levels between normal and Etnk2 Ϫ/Ϫ mice (24.8 Ϯ 7.0 compared with 26.0 Ϯ 6.2 ng/ml serum). These data illustrate that the hormonal status of the Etnk2 Ϫ/Ϫ females was the same as that of the wild-type females and does not account for the reproductive phenotype.
At embryonic day (E) 7.5, embryos from Etnk2 Ϫ/Ϫ and control mice were dissected, and transverse sections were stained (Fig. 7A). Both placentas and embryos were at a normal stage of development compared with the wild-type controls. There was no reduction in litter size, and there were no resorbed or empty conceptuses in the Etnk2 Ϫ/Ϫ mice. All embryos had similar sized embryonic cavities and decidual swelling, and no appar-FIGURE 5. Analysis of PtdEtn and PtdCho molecular species in wild-type and Etnk2 ؊/؊ mice. Livers were removed from 3-month-old mice maintained on a standard laboratory diet; the lipids were extracted; and the mass spectra were obtained as described under "Experimental Procedures." The fatty acid chains that constitute the major molecular species are indicated in shorthand for the number of carbon atom/number of double bonds. A and B, the PtdEtn molecular species profile in wild-type and Etnk2 Ϫ/Ϫ mice, respectively; C and D, the PtdCho molecular species distribution in wild-type and Etnk2 Ϫ/Ϫ mice, respectively. SEPTEMBER 22, 2006 • VOLUME 281 • NUMBER 38 ent growth retardation was observed. Sections from Etnk2 Ϫ/Ϫ embryos indicated that development was in the process of gastrulation and that all three germ layers were formed and had normal extraembryonic membranes. The morphology, number, and distribution of giant cells appeared normal in the growing placentas. The extraembryonic ectodermal sheet of the chorion was also similar to that in wild-type embryos. These data suggest that implantation and initial growth of placentas were not defective in female Etnk2 Ϫ/Ϫ mice. The morphology and number of embryos found at E12.5 were also apparently normal (Fig. 7B), indicating that fetal development was not arrested at this point. At E12.5, the formation of the chorioallantoic connection and the structures of the spongiotrophoblast layer and labyrinth layer in Etnk2 Ϫ/Ϫ mice appeared normal in comparison with those in wild-type mice (Fig. 7C). No significant mor-

Etnk2 Knock-out Mice
phological differences in the littermates of pregnant Etnk2 Ϫ/Ϫ females were observed at this stage of development. However, resorbed embryos appeared at an ϳ1:3 ratio to normally developed embryos at E17.5 in Etnk2 Ϫ/Ϫ mice (Fig. 8A). The resorbed embryos were smaller and yellow or orange in comparison with the controls, and placenta fibrosis was evident. These observations correlated with reduction of the litter size in Etnk2 Ϫ/Ϫ mice and demonstrated that Etnk2 plays a nonredundant role in the maternal support of embryonic development during the late stages of pregnancy. Placentas and embryos were isolated from pregnant Etnk2 Ϫ/Ϫ and wild-type control mice at E17.5, and morpholog-ical analysis was performed on four viable and nonviable littermate embryos. Large blood clots were clearly visible in the labyrinth layer of the placentas of nonviable embryos at E17.5 (Fig. 8B, right  panel), but not in those of normal littermates (Fig. 8B, left panel). Approximately 20% of the total volume of the labyrinth later was occupied by thrombi in the nonviable embryos, and all of the nonviable Etnk2 Ϫ/Ϫ placentas showed this striking abnormality. The viable embryos exhibited variable amounts of thrombosis. These thrombotic events were smaller, fewer in number, and located primarily near the edge of the labyrinth layer. Small thrombi occasionally formed in the placentas of wild-type embryos; however, the prevalence and size of thrombi in the Etnk2 Ϫ/Ϫ placentas clearly distinguished them from normal mouse placentas. The tissue structures of the maternal decidual layer, chorionic plate, and spongiotrophoblast layer remained largely unaltered in Etnk2 Ϫ/Ϫ mice compared with wild-type females. The functional consequence of the thrombus formation in placentas from Etnk2 Ϫ/Ϫ mice, determined by using the TUNEL assay for apoptosis (Fig.  8C) and immunohistochemistry to detect activated caspase-3 (Fig. 8D), revealed clear differences between the nonviable and viable Etnk2 Ϫ/Ϫ placentas. Extensive staining of the epithelial cells in the labyrinth layer was observed in nonviable embryos, whereas staining of the spongiotrophoblast layer and chorionic plate was less pronounced. Little staining with either reagent was observed in viable Etnk2 Ϫ/Ϫ placentas (Fig. 8C, left panel) and in wild-type embryos (data not shown). These findings indicate that the placental abnormalities observed in Etnk2 Ϫ/Ϫ mice were caused by thrombus formation, leading to restriction of the blood supply to the embryo, extensive apoptosis, and embryonic death.

DISCUSSION
This work has identified the second mammalian Etn-specific kinase that is most highly expressed in liver. Like EtnK1 (28), the EtnK2 enzyme is highly selective for Etn and phosphorylates Cho poorly, if at all. The livers of Etnk2 Ϫ/Ϫ animals are clearly deficient in EtnK activity, showing that

Etnk2 Knock-out Mice
EtnK2 is the major protein responsible for Etn phosphorylation in this tissue. In light of the fact that EtnK2 is clearly the predominate kinase in the CDP-Etn pathway in liver, the finding that hepatocytes are only slightly compromised in the incorporation of exogenous Etn into PtdEtn and that phospholipid content and structure are the same in the wildtype and Etnk2 Ϫ/Ϫ livers is perhaps surprising. Cho deprivation is a nutritional scenario that places a heavy burden on liver PtdEtn synthesis to produce PtdCho (26,27), but placing the Etnk2 Ϫ/Ϫ mice on a Cho-deficient diet did not have a discernable effect on liver phospholipid composition or pathology. This is in stark contrast to mice lacking PtdEtn N-methyltransferase expression, where liver dysfunction and pathology become apparent after only a few days (11,26,27). A deficiency in phosphocholine cytidylyltransferase (29) and PtdEtn N-methyltransferase (30) also has only a modest effect on hepatic lipid metabolism. These data, together with our analysis of Etnk2 Ϫ/Ϫ mice, serve to point out the robust nature of liver phospholipid metabolism.
The brains and testes of Etnk2 Ϫ/Ϫ mice appear normal, contrary to predictions from the study of other systems and the expression pattern of Etnk2. The Drosophila eas mutant is a null allele for EtnK that gives rise to flies that become paralyzed when vortexed (10,31). These flies are not grossly deficient in PtdEtn, but they do have slightly lower levels of PtdEtn in the brain, and their phenotype is thought to arise from an excitability defect caused by altered local membrane phospholipid composition. Although mouse EtnK2 is a homolog of the eas gene product, we did not detect any overt neurological defects in our animals. In contrast to Drosophila, where eas expression is the highest in neural tissue, mice do not express a large amount of the Etnk2 transcript in the brain. EtnK1 is the more prominent brain isozyme, and Etnk2 gene inactivation does not appreciably lower total brain tissue EtnK activity. The finding that Etnk2 expression is specifically up-regulated in Sertoli's cells of the testis during mouse sex determination (13) suggests that the Etnk2 Ϫ/Ϫ knock-out mice would exhibit defects in testicular development. However, we were unable to detect a morphological abnormality or reproductive failure of Etnk2 Ϫ/Ϫ testis. These considerations led to the conclusion that mammalian tissues have considerable plasticity with regard to PtdEtn synthesis, which is supported by Etn phosphorylation by EtnK1, EtnK2, and the ChoKs and through the operation of the separate phosphatidylserine decarboxylase pathway. Sphingosine 1-phosphate degradation should be considered as another potential source for phosphoethanolamine (32).
The maternal-specific failure to support late embryonic development in Etnk2 Ϫ/Ϫ mice reveals a non-redundant role for EtnK2 in placental function. Etnk2 Ϫ/Ϫ females give birth to smaller litters, with a mixture of average size pups and low birth weight animals, which often die within 1 week after birth, and a percentage of the embryos die in utero late in gestation. The observation of placental thrombosis and apoptosis associated with the aborted embryos points to the inability of pregnant Etnk2 Ϫ/Ϫ mice to control coagulation in the placental labyrinth, leading to intrauterine growth retardation and/or death. The maternal components of the placenta derive from the maternal vasculature and uterine decidual cells, which come into close contact with the embryonic cells of the trophoblast lineage in the labyrinth layer (33). There is a delicate balance between clotting and hemorrhage in the placenta (34), and there are multiple factors that impact on maintaining this balance. The estrogen sulfatase knock-out mouse has a similar maternal-specific placental thrombosis phenotype because of the elevation of plasma estradiol (35); however, the comparable estrogen levels in Etnk2 Ϫ/Ϫ and control mice rule out a contribution of estradiol dysregulation to our phenotype. The ROCK-II knock-out phenotype of intrauterine growth retardation and thrombosis is attributed to the inability of embryonic trophoblasts to control coagulation (16). Although this phenotype is similar to the Etnk2 Ϫ/Ϫ phenotype, it differs in being fetal-specific rather than maternal-specific. However, the processes of cell-surface PtdEtn movement and RhoA signaling were recently linked in cultured cells (19); therefore, it is possible that activation of the Rho/ROCK signaling pathway in embryonic trophoblasts may be defective because of the lack of extracellular maternally derived PtdEtn. The protein C anticoagulation pathway is established as critical for placental function (36). A deficiency in protein C gives rise to placental thrombosis and intrauterine fetal death (37) and increases apoptosis in hypoxic endothelial cells (38). PtdEtn is a critical cofactor in the protein C activation pathway (39), raising the possibility that reduced PtdEtn in the placentas of Etnk2 Ϫ/Ϫ females may compromise the generation of activated protein C, leading to thrombosis and hypoxic apoptosis in the placenta. These data indicate the importance of EtnK2 expression and, by extension, maternally derived PtdEtn in maintaining hemostasis in the placental labyrinth. Recurrent pregnancy loss and thrombosis are also characteristic of the antiphospholipid syndrome, a multifaceted collection of syndromes associated with antibodies against specific lipid-protein combinations (40). One class of anti-phospholipid antibodies inhibits protein C activation in a PtdEtn-dependent manner (41). The lack of maternal PtdEtn or maternal PtdEtn-protein C or PtdEtn-␤ 2glycoprotein I complexes may create an imbalance in the regulation of coagulation by the embryonic trophoblasts lining the blood vessels. Thus, the hypothesis that EtnK2 has a non-redundant role in producing endothelial cell PtdEtn to support activation of the protein C anticoagulation cascade appears a tenable direction for future work.