Specificity and rate of human and mouse liver and plasma phosphatidylcholine synthesis analyzed in vivo.

Phosphatidylcholine (PC) synthesis by the direct cytidine diphosphate choline (CDP-choline) pathway in rat liver generates predominantly mono- and di-unsaturated molecular species, while polyunsaturated PC species are synthesized largely by the phosphatidylethanolamine-N-methyltransferase (PEMT) pathway. Although altered PC synthesis has been suggested to contribute to development of hepatocarcinoma and nonalcoholic steatohepatitis, analysis of the specificity of hepatic PC metabolism in human patients has been limited by the lack of sensitive and safe methodologies. Here we incorporated a deuterated methyl-d9-labled choline chloride, to quantify biosynthesis fluxes through both of the PC synthetic pathways in vivo in human volunteers and compared these fluxes with those in mice. Rates and molecular specificities of label incorporated into mouse liver and plasma PC were very similar and strongly suggest that label incorporation into human plasma PC can provide a direct measure of hepatic PC synthesis in human subjects. Importantly, we demonstrate for the first time that the PEMT pathway in human liver is selective for polyunsaturated PC species, especially those containing docosahexaenoic acid. Finally, we present a multiple isotopomer distribution analysis approach, based on transfer of deuterated methyl groups to S-adenosylmethionine and subsequent sequential methylations of PE, to quantify absolute flux rates through the PEMT pathway that are applicable to studies of liver dysfunction in clinical studies.

instance, PEMT expression and activity are low both in human hepatocellular carcinomas ( 16 ) and during rat liver growth after partial hepatectomy ( 17 ).
PC synthesis in human liver disease has received less attention largely because of the lack of appropriately safe methodologies. Studies using a combination of isotope ratio mass spectrometry (MS) and multiple isotopomer distribution analysis (MIDA) modeling have determined rates of hepatic triacylglycerol synthesis and VLDL secretion by using stable isotopes (reviewed in [ 18 ]) but have not investigated the synthesis and turnover of hepatic or plasma PC. An alternative analytical approach ( 19 ) analyzes incorporation of stable isotope-labeled substrates into intact PC molecules. Incubation of rat hepatocytes with deuterated methyl-D 9 -labeled choline chloride, followed by electrospray ionization tandem MS (ESI-MS/ MS), enabled determination of the synthesis of individual PC molecular species. Moreover, choline is a major methyl donor for synthesis of SAMe, resulting in transfer of a labeled methyl group to form methyl-D 3 -SAMe ( Fig. 1B ) ( 19 ). pathway forms mainly monounsaturated and di-unsaturated PC species such as palmitoyloleoyl-PC (PC16:0/18:1) and palmitoyl linoleoyl-PC (PC16:0/18:2), whereas the PE-N methylation pathway synthesizes PUFA-containing PC species such as palmitoyldocosahexaenoyl-PC (PC16:0/ 22:6) and palmitoylarachidonoyl-PC (PC16:0/20:4) (in this molecular species nomenclature, e.g., A:a/B:b, A and B represent the total number of carbon atoms in each fatty acid at the sn -1 and sn -2 positions of the glycerophosphate backbone, respectively, and a and b represent the number of unsaturated double bonds in each fatty acid). The importance of PE-N methylation for the supply of PUFA from the liver to peripheral tissues is highlighted in PEMT Ϫ / Ϫ mice, which have substantially diminished plasma concentrations of docosahexaenoyl-containing (22:6 n-3 ) and arachidonoyl-containing (20:4 n-6 ) PC species, although fatal liver failure is prevented and total plasma PC is normalized by dietary choline supplementation ( 13,14 ). Considerable evidence suggests that PE-N methylation in liver is involved in the regulation of cell division ( 15 ). For

Analysis of plasma and tissue phospholipids and tissue choline phosphate
A total of 100 mg of mouse liver or lung was homogenized (Ultra-Turrax; Janke and Kunkel, Germany) in 3 ml of 0.9% saline on ice. Lipids were extracted from homogenate (800 l) and plasma (50 l) by using chloroform and methanol ( 21 ), dried under nitrogen gas, and analyzed by ESI-MS. Dimyristoyl PC (PC14:0/14:0) and dimyristoyl PE (PE14:0/14:0) were added as internal recovery quantifi cation standards. PC species were quantifi ed using a Waters Micromass Quattro Ultima triple quadrupole mass spectrometer (Micromass, Wythenshaw, UK) equipped with an ESI interface. Samples were dissolved in methanol-chloroform-water-25% NH 4 OH (7:2:0.8:0.2,v/v) and introduced into the mass spectrometer by direct infusion. Collision gas-induced dissociation using argon gas produced protonated phosphocholine head group fragments with m/z of +184, corresponding to endogenous PC, and m/z of +193 for PC synthesized via the CDPcholine pathway ( 22 ). Analysis of the products with m/z +187 or +190 (i.e., containing one or two recycled deuterium-labeled methyl groups) revealed indirect incorporation of deuterium into the PC head group via the PEMT pathway ( 23 ). PC molecular species were quantifi ed from diagnostic precursor ion scans of the respective fragment ions (P184, P187, P190, P193). A diagnostic neutral loss scan of m/z +141 (NL141) provided analysis of diacyl PE molecular species. Concentrations of infused phospholipids were optimized within the linear response range of the mass spectrometer. Importantly, varying the concentration of infused lipid within this range did not signifi cantly affect the PC or PE molecular species profi les or the fractional incorporation of stable isotope labels. Phospholipid spectra were quantifi ed using MassLynx software and Excel spreadsheets (see Supplementary material).
The aqueous upper phase from tissue extracts was collected and lyophilized, and enrichment of the methyl -D 9 label in substrate choline phosphate was determined. Choline phosphate was isolated using aminopropyl solid phase cartridges (Varian Inc., Palo Alto, CA) ( 24 ). Endogenous and methyl -D 9

ESI-MS/MS analysis of molecular specifi cities of PC synthesis pathways in mouse liver
Incorporation patterns of labeled precursors into mouse liver PC by both the CDP-choline and the PEMT pathways 1.5 h after injection of methyl-D 9 -choline chloride are illustrated in Fig. 2 . Spectra have been mass displaced on the x-axis to align PC, newly synthesized PC, and PE with equivalent molecular species components. They show the almost complete segregation of molecular specifi cities of PC synthesis by the two pathways. Endogenous PC ( Fig.  2A , P184 MS scan) was a combination of mono-, di-, and polyunsaturated species. PC synthesized directly by the CDP-choline pathway ( Fig. 2B , P193 scan) was preferentially enriched in mono-(PC16:0/18:1) and di-unsaturated (PC16:0/18:2) species. In contrast, PC synthesized by the PEMT pathway ( Fig. 2C , P187, containing one recycled methyl -D 3 group) was predominantly polyunsaturated, and PC16:0/22:6 was the major species formed. This pattern of molecular species was similar but not identical to that of Consequently, this method is also used to monitor the PEMT pathway fl ux from the incorporation of the methyl -D 3 label ( Fig. 1B ), as well as to direct incorporation of the methyl -D 9 label. Importantly, these analyses of PC biosynthesis fl uxes provide information about the individual molecular species that are biologically relevant molecules, and, consequently, analyses of their kinetics are essential to determine the contribution of altered PC metabolism to disease mechanisms.
We previously applied ESI-MS/MS methodology to quantify the synthesis and turnover of individual molecular species of PC in lung surfactant in human volunteers ( 20 ) and now use it to describe the synthesis of individual molecular species of hepatic PC in vivo. Initial analyses were conducted in a mouse model to establish whether analysis of the molecular specifi city and dynamics of label incorporation into plasma PC can provide a direct insight into liver PC metabolism. This methodology was then applied to human volunteers to show for the fi rst time the molecular specifi city of PC synthesis by human liver in vivo, monitored by the appearance of label in plasma PC. Finally, we present a mass isotopomer analysis for the calculation of substrate fl ux through the PEMT pathway that highlights both the practicality and diagnostic/investigative potential of stable isotope labeling with methyl-D 9 -choline for the study of diseases involving altered hepatic lipid metabolism in humans in vivo.

Animal labeling
Eight-week-old C57BL6 female mice (20.8 ± 1.33 g) received intraperitoneal (ip) injections of 100 l of methyl -D 9 -choline chloride (10 mg/ml in sterile physiological saline) and were then sacrifi ced at 1.5, 3, 6, and 24 h ( n = 5 mice per time point). Blood was drawn by heart puncture and collected in Microtainer® lithium heparin tubes (Becton Dickinson, Oxford, UK), and plasma was isolated by centrifugation (1,000 g for 10 min) and stored at Ϫ 20°C. Livers and lungs were removed, snap frozen in liquid nitrogen, and stored at -80°C. Animals received appropriate humane care according to National Institutes of Health guidelines. All animal procedures were conducted under United Kingdom Home Offi ce Animal Procedures License PPL30/1661.

Human labeling
Methyl -D 9 -choline chloride (3.6 mg/kg of body weight) was infused intravenously over 3 h in 1 male and 3 healthy human female volunteers (22-27 years old; weighing 63-92 kg). Venous blood samples mixed with EDTA were collected at intervals from 6 to 168 h after the start of methyl-D 9 -choline chloride infusion; samples were centrifuged at 1,000 g at 4°C for 15min, and plasma was stored at -20°C. Initial blood samples and those collected at 24, 48, 72, 96, and 168 h were taken after overnight fasting, but volunteers did not fast before all other samples were taken. All volunteers ate their normal diets and consumed a light breakfast before the infusions were started. There was no attempt in this pilot study to control diets before or during the experimental days. The study was approved by the local ethics committee at Hannover Medical School, subjects' written consent was obtained, and standard laboratory parameters showed no signs of impaired renal or hepatic function. ately less PC being synthesized by the PE-N methylation pathway in human than in mouse liver.

Molecular specifi city of PC synthesis by the CDP-choline and PE-N methylation pathways
The composition of endogenous and newly synthesized PC is detailed in Supplementary Table S1 and summarized in Fig. 4 for the most informative PC species. These species were selected to illustrate the major alterations that occurred over time to the stable isotope incorporation patterns by the CDP-choline ( Fig. 4A, C, E ) and PE-N methylation ( Fig. 4B, D, F ) pathways into PC from mouse liver ( Fig. 4A, B ), mouse plasma ( Fig. 4CD ), and human plasma ( Fig. 4E, F ). These changes are remarkably consistent across different sample types, although the magnitude of the effect differs. Temporal alteration to mouse liver PC synthesis by the CDP-choline pathway is consistent with acyl remodeling at both the sn -1 and the sn -2 positions, with initial synthesis as PC16:0/18:2, followed by progressive redistribution of label to PC18:0/18:2 and PC18:0/20:4, and resembles the endogenous PC composition by 24 h ( Fig. 4A ). The equivalent labeling pattern is essentially identical for mouse plasma ( Fig. 4C ) and, importantly, very similar for human plasma ( Fig. 4E ).
Initial incorporation into PC16:0/22:6 in mouse liver PC by the PE-N methylation pathway (

Kinetics of incorporation of methyl -D 9 -choline into PC by the CDP-choline and PE-N methylation pathways
The totals of all PC species shown in Fig. 2B with abundance proportions greater than 1% of the largest ion peak in each spectrum were quantifi ed for P184, P187, P190, and P183 MS scans in extracts of mouse liver, lungs, and blood plasma and extracts of human plasma. Lung was chosen for comparison, as another organ specialized for PC secretion, in this case, pulmonary surfactant. The fractional incorporation of methyl-D 9 label into mouse liver and lung PC and expressed relative to total PC is shown in Fig. 3A, and the corresponding incorporation of methyl -D 3 label is shown in Fig. 3B . The initial rates of hepatic and lung PC methyl-D 9 labeling were already very high by 1.5 h, reached maxima at 3 h, and subsequently declined rapidly. Fig. 3B shows methyl -D 3 labeling was almost exclusive for liver PC at 1.5, 3, and 6 h; the lung labeling at 24 h represented PC synthesis by the CDP-choline pathway from recycled methyl -D 3 -choline and was not associated with PE-N methylation (Supplementary Fig. S1).
Appearance of label in mouse plasma PC closely refl ected labeling of liver PC by both pathways ( Fig. 3C ), demonstrating that not only was PC synthesized very rapidly by mouse liver but that it was secreted equally rapidly into the circulation. The comparable pattern of stable isotope labeling of human plasma PC over a more prolonged time scale ( Fig. 3D ) demonstrated that both pathways were active in the synthesis of lipoprotein PC, with proportion-

Rate of mouse liver PC synthesized by the CDP-choline pathway
The maximal rate of PC synthesis by the CDP-choline pathway at 3 h was calculated based on the assumptions that 1 ) all methyl -D 9 label was incorporated during the previous 1.5 h; 2 ) PC labeled between 0 and 1.5 h was already secreted; 3 ) enrichment of the choline phosphate substrate pool could be accurately estimated; and 4 ) the whole system was at steady state. The fractional enrichment of methyl -D 9 -choline phosphate ( Fig. 6A ) in liver followed a nonlinear distribution, with a rapid initial decline to 3 h, followed by a slower decrease to 24 h. Although these data are limited, they suggest a biphasic response, with the latter period representing uptake of lipoprotein by the liver and recycling of labeled choline ( 25 ). Given that at 3 h the mean enrichment of methyl -D 9 -PC was 4.15% and that of methyl -D 9 -choline phosphate was 5.97% between 1.5 and 3 h, an estimate of the maximal rate of synthesis by this pathway was calculated as (4.15 × 100)/(5.97 × 1.5) = 46.3% of total liver PC/h.

Rate of mouse and human liver PC synthesis by PE-N methylation
A similar estimate for PEMT pathway fl ux was calculated using MIDA to determine the fractional enrichment of the essentially mirrored that in liver PC ( Fig. 4D ). The corresponding distribution of methyl -D 3 label in human plasma PC molecular species followed a very similar pattern, again with an initial incorporation into PC16:0/22:6 and PC18:0/20:4 and a subsequent redistribution to 18:2containing species ( Fig. 4F ). This analysis demonstrates for the fi rst time that the PEMT pathway in human liver is highly specifi c for the synthesis of 22:6-and 20:4-containing PC molecular species destined for lipoprotein secretion.
Importantly, the similar labeling patterns of mouse liver and plasma PC by both pathways at each time point clearly show that, although newly synthesized liver PC undergoes a substantial process of acyl remodeling after its initial synthesis, this is not required for secretion of lipoprotein PC. Instead, these results suggest that newly synthesized PC is secreted from the liver into the circulation with minimal modifi cation to the initial synthesis pattern. While liver PC labeling data are not available, the human plasma labeling results also strongly suggest a similar mechanism for synthesis and secretion of human plasma lipoprotein PC.

Turnover of individual PC species synthesized by PE-N methylation
The methyl -D 3 fractional incorporations illustrated in Fig. 3 represent the mean turnovers of all selected PC species, but Fig. 4 shows how the distribution of incorporated label into two selected PC species changes with time. PC16:0/22:6 and PC16:0/18:2 are the characteristic PC species synthesized, respectively, by the PE-N methylation Fig. 3. Fractional enrichment of methyl -D 9 and methyl -D 3 labels in liver, lung, or plasma total PC in vivo. A: Direct incorporation of methyl -D 9 label by the CDP-choline pathway into mouse liver (circles) and lung (squares) total PC over a 24 h period after ip injection of 1 mg of methyl -D 9 -choline chloride ( n = 5 mice/ group). B: Incorporation of methyl -D 3 label in SAMe mice, demonstrating the absence of the PEMT pathway in mouse lung. C and D: Time courses of incorporation of the methyl -D 9 (squares) and methyl -D 3 (circles) labels into, respectively, mouse an human plasma. Results were calculated as a percentage of total PC, given by the sum of P184, P187, P190, and P193 for all selected species (means ± SD). ments to provide estimates of total PC synthesis by PE-N methylation at each time point for mouse liver and plasma ( Fig. 7A ) and human plasma ( Fig. 7B ). Initial incorporation phases for each plot were linear, providing estimates of fl ux though this pathway. The similarity in synthetic rates for mouse liver (5.04%/h) and plasma (4.71%/h) strongly suggests that the value of 0.53%/h calculated for human plasma can be used as a surrogate for human liver PC fl ux by the PEMT pathway in vivo. The specifi city of this synthesis ( Fig. 8 ) for fi ve representative PC species (PC16:0/22:6, PC18:0/22:6, PC18:0/ 20:4, PC16:0/18:2, and PC16:0/18:1) showed that synthesis rates varied over 1 order of magnitude for all three sample types, with essentially identical rank orders, and was greatest for 22:6-containing PC species, intermediate for PC18:0/20:4, and lowest for PC16:0/18:1 and PC16:0/18:2. This similarity in synthetic rates further suggests that this stable isotope labeling approach can be used to probe mechanisms regulating the specifi city of hepatic PEMT activity in vivo in human subjects as well as in animals experiments. methyl -D 3 -SAMe pool in the liver ( 26 ). Under these experimental conditions, PE can incorporate up to three methyl -D 3 groups during its conversion to PC by PEMT. Incorporation of one methyl -D 3 group will be detected by a P187 MS scan and that of two methyl -D 3 groups by a P190 scan, with comparable patterns of incorporation (Supplementary Fig. S2). Enrichment of methyl -D 3 -SAMe is then calculated with the equation P190/ ⌺ (P187 + P190) (see Supplementary Equation S3).
Importantly for clinical applications, very similar enrichment values can be calculated from the analysis of both mouse plasma and liver PC ( Fig. 6B ). By analogy, an equivalent calculation can then be proposed using human plasma PC data to provide a surrogate measure of the methyl -D 3 -SAMe enrichment in human liver ( Fig. 6C ).
These values for methyl -D 3 -SAMe enrichment were then used to correct corresponding methyl-D 3 -PC enrich-  This ability to quantify PEMT metabolic fl ux in vivo has considerable potential clinical application to a wide range of metabolic and hepatic diseases. Liver PEMT is a significant methylation activity in the body ( 27 ) and is a principal determinant of plasma homocysteine concentration ( 28,29 ), an established risk factor for coronary heart disease ( 30 ). Activity of PEMT is decreased in liver biopsies from patients with cirrhosis ( 31 ); the PEMPT pathway lossof-function V175M polymorphism is associated with NAFL ( 3,32 ); PEMT Ϫ / Ϫ mice develop fatal steatohepatitis when placed on a choline-defi cient diet ( 33 ); and decreased PEMT pathway fl ux is an essential component of the choline-methionine-defi cient dietary mouse model of NAFL ( 34 ). PEMT Ϫ / Ϫ mice on a choline-suffi cient diet have a normal liver PC content but decreased concentration of plasma PC and PC species containing 20:4 and 22:6 ( 13 ). Results shown in Figs. 4 and 8 highlight the importance of PEMT activity in hepatic lipoprotein exportation of PUFAcontaining PC to extrahepatic tissues and strongly suggest that virtually all newly synthesized PUFA-containing PC species secreted by mouse liver are derived from PE-N methylation and not from the CDP-choline pathway. Equivalent distributions and kinetic changes of methyl -D 9 and methyl -D 3 labels in human compared with mouse plasma PC suggest similar mechanisms for the metabolism and secretion of lipoprotein PC in humans, although this remains to be confi rmed in future studies using human liver biopsies. A small proportion of VLDL is synthesized by enterocytes, but, while de novo synthesis from choline DISCUSSION Incorporation of methyl -D 9 -labeled choline has previously been used to characterize the molecular specifi city of PC synthesis by isolated rat hepatocytes and hepatoma cells, but the results presented here are the fi rst demonstration of the specifi city of human plasma PC synthesis. Strong similarities between the specifi cities of mouse liver and plasma PC synthesis further suggest that incorporation of label into human plasma PC can be used as a surrogate for human liver PC synthesis in vivo. This experimental approach is relatively inexpensive, rapid, and straightforward and eliminates the use of substrates labeled with radioactive isotopes. In particular, use of MIDA for estimating methyl -D 3 -SAMe enrichment enables absolute rates of synthetic fl ux through the PEMT pathway to be estimated, in turn providing a noninvasive method for determining liver PEMT activity in vivo in patients. It is important to emphasize that quantifi cation of methyl -D 3 -SAMe enrichment by this experimental approach is independent of the initial source of the methyl donor and would be equally applicable, for instance, to studies of the incorporation of methyl -D 3 -methionine.  eling before being secreted. As acyl remodeling apparently occurs over a period of 3-6 h ( 9, 11 ) and as our estimate of the combined fl ux rate through both pathways is approximately 50%/h for both mouse liver and plasma PC, PC acyl remodeling must be only one of a number of activities that determine liver PC molecular species composition. In this context, the relative contributions to the equilibrium composition of liver and plasma PC of LPCAT ( 37 ), lecithin-cholesterol acyltransferase ( 38 ), recycling of extrahepatic PC to the liver ( 25 ), and secretion of PC in bile ( 39 ) have yet to be fully evaluated.
Our data have strongly suggested that modulation of the N methylation pathway provides capacity for the directed supply of PUFA from the liver to extrahepatic tissues, which is largely independent of PC synthesis by the CDP-choline pathway. Use of MIDA to calculate absolute rates of PC synthesis by PE-N methylation enables this pathway fl ux to be calculated in vivo in human patients and consequently has considerable potential for the investigation of a range of important liver diseases, including hepatic cancers, alcoholic and nonalcoholic fatty liver disease, and cystic fi brosis.
is essential for PC secretion into the intestinal lumen ( 35 ), the contribution of de novo PC synthesis to enterocyte plasma lipoprotein packaging and secretion is not known. The methodology presented here does not directly address this potential enterocyte contribution.
The similarity of molecular specifi cities of newly synthesized PC synthesis in mouse liver and plasma has significant implications for mechanisms regulating assembly and secretion of plasma lipoproteins. Previous studies have shown that rat liver PC ( 9,11 ) is synthesized by the CDPcholine pathway largely as PC16:0/18:2 and PC 16:0/18:1 and is then acyl remodeled to the mature composition by a variety of phospholipase and lysophosphatidylcholine acyltransferase (LPCAT) activities ( 36,37 ). Results presented in Fig. 4 clearly demonstrate that liver PC, newly synthesized by either pathway, undergoes little acyl remod-