Small Intestine but Not Liver Lysophosphatidylcholine Acyltransferase 3 (Lpcat3) Deficiency Has a Dominant Effect on Plasma Lipid Metabolism*

Lysophosphatidylcholine acyltransferase 3 (Lpcat3) is involved in phosphatidylcholine remodeling in the small intestine and liver. We investigated lipid metabolism in inducible intestine-specific and liver-specific Lpcat3 gene knock-out mice. We produced Lpcat3-Flox/villin-Cre-ERT2 mice, which were treated with tamoxifen (at days 1, 3, 5, and 7), to delete Lpcat3 specifically in the small intestine. At day 9 after the treatment, we found that Lpcat3 deficiency in enterocytes significantly reduced polyunsaturated phosphatidylcholines in the enterocyte plasma membrane and reduced Niemann-Pick C1-like 1 (NPC1L1), CD36, ATP-binding cassette transporter 1 (ABCA1), and ABCG8 levels on the membrane, thus significantly reducing lipid absorption, cholesterol secretion through apoB-dependent and apoB-independent pathways, and plasma triglyceride, cholesterol, and phospholipid levels, as well as body weight. Moreover, Lpcat3 deficiency does not cause significant lipid accumulation in the small intestine. We also utilized adenovirus-associated virus-Cre to deplete Lpcat3 in the liver. We found that liver deficiency only reduces plasma triglyceride levels but not other lipid levels. Furthermore, there is no significant lipid accumulation in the liver. Importantly, small intestine Lpcat3 deficiency has a much bigger effect on plasma lipid levels than that of liver deficiency. Thus, inhibition of small intestine Lpcat3 might constitute a novel approach for treating hyperlipidemia.

The majority of lipids on the cell membrane as well as the plasma lipoproteins are phospholipids, in particular phosphatidylcholines (PCs) 2 (1,2). Monounsaturated and saturated fatty acids are usually esterified at the sn-1 position of PCs, whereas polyunsaturated fatty acids are esterified at the sn-2 position (3). The asymmetric distribution of fatty acids at the sn-1 and sn-2 positions of PCs is maintained in part by a deacylationreacylation process known as the Lands cycle or PC remodeling (3,4). One key enzyme in this remodeling is lysophosphatidylcholine acyltransferase (Lpcat), which utilizes lyso-PC and polyunsaturated acyl-CoA to produce sn-2 polyunsaturated PCs. There are four isoforms of this enzyme (5), and Lpcat3 is the major isoform in the small intestine and liver (6 -8).
Lpcat3 is one of the downstream targets of liver X receptor (8) and peroxisome proliferator-activated receptor ␣ (6). Acute knockdown of Lpcat3 expression in the liver of genetically obese mice exacerbates lipid-induced endoplasmic reticulum (ER) stress (8). Moreover, global Lpcat3 knock-out (KO) mice exhibit neonatal lethality and have an abnormal small intestine (9,10). Liver-specific deletion of Lpcat3 has no effect on ER stress but results in a reduction of plasma triglycerides and induction of hepatosteatosis under high fat feeding conditions (9). Because of neonatal lethality (non-inducible approach was utilized), only 1-week-old newborns were analyzed for the impact of intestine-specific Lpcat3 deficiency on lipid metabolism (9). Very recently, we found that the neonatal lethality of global Lpcat3 KO mice could be rescued by oral administration of PC plus olive oil. However, the KO mice had shorter and wider small intestinal villi and a longer and larger small intestine (11). Thus, the following remains to be investigated: 1) the impact of small intestine Lpcat3 deficiency in adult mice on lipid metabolism; and 2) the respective contribution of small intestine and liver Lpcat3 to plasma lipid levels.
In this study, we specifically ablated Lpcat3 in the small intestine of mice using an inducible Cre-LoxP approach. Our results indicate that this ablation reduces the levels of plasma cholesterol, triglycerides, and phospholipids. We also evaluate the effect of liver-specific Lpcat3 deficiency on plasma lipid metabolism. We found that small intestine but not liver Lpcat3 deficiency has a dominant effect on plasma lipid metabolism.

Materials and Methods
Generation of Intestine-specific Lpcat3-deficient Mice-Lpcat3-Flox mice (prepared by inGenious Targeting Laboratory) were crossed with Villin-Cre-ER T2 transgenic mice. A 9-kb segment of the regulatory region of the mouse Villin gene drives expression of Cre recombinase fused to a mutated ligand-binding domain of the human estrogen receptor. Cre recombination is inducible with tamoxifen treatment in epithelial cells all along the crypt-villus axis and in undifferentiated progenitor cells in the crypt region. We established Lpcat3-Flox/Villin-Cre-ER T2 mice according to the strategy shown in Fig. 1A. The genotype was confirmed by polymerase chain reaction (PCR) (Fig. 1B). To delete the gene in the small intestine, tamoxifen (2 mg/mouse, dissolved in 200 l of corn oil (Sigma C8267)) was injected intraperitoneally on four alternate days. Lpcat3-Flox mice injected with tamoxifen were used as controls. Both male and female 12-week-old mice on a C57BL6J background were used in this study. All studies were approved by the Institutional Animal Care and Use Committee of State University of New York Downstate Medical Center and conformed with the "Guide for the Care and Use of Laboratory Animals" published by the National Institutes of Health (publication 85-23, revised in 1996).
Liver-specific Lpcat3-deficient Mouse Preparation-We injected (intraperitoneally) AAV-Cre (1 ϫ 10 12 ) into the Lpca3-Flox mice. Because the Cre recombinase expression was driven by the liver-specific thyroxine-binding globulin promoter (12), the AAV approach resulted in sustained Cre expression in the liver for at least 20 weeks (13).
mRNA Measurement-Mice were sacrificed by cervical dislocation. The jejunum and liver were dissected, and total RNA was extracted from each tissue using the RNeasy mini kit (Qiagen). cDNA was synthesized with the Superscript TM III firststrand synthesis kit (Invitrogen). PCR was performed in a total volume of 20 l with the SYBR Green kit from Applied Biosystems. The amplification program was as follows: activation at 95°C for 10 min followed by 40 amplification cycles of 95°C for 15 s and 60°C for 60 s. Each sample was assayed in triplicate. The genes encoding 18S rRNA and GAPDH were used as internal controls. Relative gene expression is expressed as the mean Ϯ S.D. Mouse Lpcat3 primers forward, TTTCTGGTT-CCGCTGCATGT, and reverse, CCGACAGAATGCACACT-CCTTC, and 18SrRNA primers forward, AGTCCCTGCCCT-TTGTACACA, and reverse, GATCCGAGGGCCTCA-CTAAAC, were used.
Lipoprotein/Apolipoprotein Analysis and Lipid Absorption-Lipoproteins and apolipoproteins (apoA-I, apoB, and apoE) in mice (both sexes) were measured as described (7). Cholesterol absorption studies were performed with male mice using the conventional fecal dual-isotope ratio method (14). Absorption of triglycerides and phospholipids was measured with male mice using our reported method (15).
Primary Enterocyte Preparation and Plasma Membrane Isolation-Primary enterocytes were isolated according to two previous reports (14,16). The plasma membrane was isolated from primary enterocytes according to our protocol (17).
Measuring Total Lpcat Activity and PC Subspecies-Total Lpcat activity was measured as described (7). PC subspecies were measured with liquid chromatography-coupled tandem mass spectrometry (LC-MS/MS) as described (7).
Hematoxylin and Eosin Staining-The small intestine was dissected out and fixed with 4% formalin. The tissue was embedded in paraffin and then sliced (5 m thick). Each slice was deparaffinized and stained with hematoxylin and eosin.
Immunohistochemical Staining-Prior to antibody staining, the sections were deparaffinized in xylene, rehydrated in a gradient series of ethanol, and then subjected to high temperature antigen retrieval in 50 mM Tris-HCl (pH 9.0) containing 1 mM EDTA. Sections were permeabilized in 50 mM Tris buffer (pH 7.4) containing 0.5% (w/v) Triton X-100 and 5% horse serum. The sections were incubated with anti-NPC1L1 or anti-villin overnight at 4°C. The sections were then incubated with fluorescently labeled secondary antibodies.
Statistical Analysis-Data are expressed as the mean Ϯ S.D. Data between two groups was analyzed by the unpaired, twotailed Student's t test. A p value of less than 0.05 was considered as significant.

Production of Intestine-specific Inducible Lpcat3 KO Mice-
To prepare intestine-specific Lpcat3-deficient mice, tamoxifen (2 mg/mouse), dissolved in corn oil, was injected intraperitoneally into female mice at days 1, 3, 5, and 7. Tamoxifen-treated Lpcat3-Flox mice were used as controls. We collected plasma and tissues on day 9 after the first treatment and measured Lpcat3 mRNA in the small intestine, liver, kidney, and adipose tissue. Compared with controls, Lpcat3 mRNA level was decreased by 90% in the small intestine but no other tissues (Fig.  1C). The levels of the remaining Lpcat isoforms (Lpcat1, -2, and -4) did not change in these Lpcat3-deficient mice (Fig. 1D). We then measured total Lpcat activity in the small intestine homogenate and found it was decreased by 85% in the Lpcat3-deficient mice compared with controls ( Fig. 1E). Moreover, we did not observe morphological changes in the small intestine of Lpcat3-deficient mice ( Fig 1F). Similar results were obtained with male mice (data not shown).
Lpcat3 Deficiency in the Small Intestine Significantly Decreases Body Weight and Plasma Lipid Levels-Lpcat3-deficient female mice underwent a significant loss of body weight compared with the controls ( Fig. 2A) at day 9 after tamoxifen treatment. We next measured fasting plasma lipid levels in the Lpcat3 KO and control female mice. Lpcat3 deficiency significantly decreased plasma levels of total cholesterol (47%, p Ͻ 0.01), total phospholipids (51%, p Ͻ 0.01), and triglycerides (81%, p Ͻ 0.01) as compared with controls ( Fig. 2, B-D). The lipid reduction was also observed at day 12 after tamoxifen treatment (Fig. 2, B-D). Similar results were obtained with male mice (data not shown).
Plasma lipid distributions in pooled plasma were examined by fast protein liquid chromatography (FPLC). Cholesterol level was dramatically decreased in both high density lipoprotein (HDL) and non-HDL fractions from the Lpcat3-deficient mice compared with controls ( Fig. 2F). Triglyceride in non-HDL was dramatically decreased in the deficient mice (Fig. 2F). Similar results were obtained with male mice (data not shown). We also measured small intestine lipids, including triglyceride, total cholesterol, phosphatidylcholine, and sphingomyelin, and we did not find significant changes (Fig. 2G).
Lpcat3 Deficiency Significantly Decreases Lipid Absorption in the Small Intestine-A defect in lipid uptake could be one of the major reasons for the observed lower plasma lipid profiles. We thus measured cholesterol absorption using a conventional fecal dual-isotope ratio method at day 9 after tamoxifen treatment. This involved the gavage of a single bolus of 0.1 Ci of [ 14 C]cholesterol and 0.2 Ci of [ 3 H]sitostanol in 20 l of olive oil, with feces being collected after 48 h. We found that the Lpcat3 KO mice absorbed significantly less cholesterol than controls (Fig. 3A). Moreover, there was much less [ 14 C]cholesterol in the plasma of Lpcat3 KO mice (Fig. 3B). We also measured triglyceride and PC absorption using [ 3 H]triolein and [ 14 C]arachidonyl-PC. The mice were gavaged with both radiolabeled lipids, and blood radioactivity was monitored 2, 4, and 8 h later. Lpcat3 deficiency significantly decreased [ 3 H]glycerolipids (Fig. 3C) and [ 14 C] glycerolipids in blood (Fig. 3D).
We hypothesized that Lpcat3 deficiency might decrease polyunsaturated PCs in the enterocyte apical membrane, resulting in diminished receptor density or membrane-associated transporter levels, thus reducing dietary lipid uptake and plasma lipid levels. We analyzed the populations of PC subspecies in the enterocyte plasma membrane using LC-MS/MS and found that Lpcat3 deficiency decreased the amount of polyunsaturated PCs (16:0/18:2, 18:1/18:2, 18:1/18:3, and 18:0/20:4) in the plasma membrane (Table 1). These changes could impact lipid uptake by the apical membrane. We also measured lyso-PCs but did not find significant changes (data not shown). We measured PC subspecies in the liver and kidney homogenates of the deficient mice and controls, and we did not find significant changes (Table 1).
We next used Western blotting to measure levels of NPC1L1, CD36, ABCA1, and ABCG8 in enterocyte homogenates; the level of each of these proteins was significantly decreased in Lpcat3-deficient primary enterocytes compared with controls (Fig. 4, A and B). We also measured MTP and FATP4 levels and did not find any significant differences in MTP. Interestingly, however, Lpcat3 deficiency increased FATP4 levels as a compensatory mechanism.
To rationalize the defects in lipid absorption, we performed immunofluorescence staining for the localization of NPC1L1 in the small intestine. NPC1L1 co-localized with the plasma membrane marker villin in the small intestine of control mice (Fig.  4C). This co-localization was greatly reduced, however, in the

Lpcat3 Deficiency and Lipid Metabolism
APRIL 1, 2016 • VOLUME 291 • NUMBER 14 small intestine of Lpcat3-deficient mice, especially at the tip of the villus. We also isolated primary enterocytes from both control and Lpcat3 KO mice and treated the cells with cycloheximide to block protein biosynthesis. Then we did Western blot for NPC1L1 at different time points. We did not find significant changes in the NPC1L1 protein decay curve within 3 h, suggesting Lpcat3 deficiency might not influence NPC1L1 protein stability (Fig. 4D). We next immuno-stained stem cells in the crypt, using leucine-rich repeat-containing G-protein-coupled receptor 5 (LGR5) antibody (18), and we did not find significant changes (Fig. 5). We also measured SREBP-1c and PPAR␣ levels, as well as their downstream target mRNA levels, and we did not find significant changes (data not shown).
Enterocytes secrete cholesterol as a part of apoB lipoproteins into lymph and as a part of HDL into the circulation (14,19). The latter is influenced by the ABCA1 level (19). Because we found that both NPC1L1 and ABCA1 levels were significantly reduced in enterocytes (Fig. 4, A and B), we next assessed the two aspects of cholesterol absorption, uptake and secretion. Primary enterocytes were incubated with [ 14 C]cholesterol for 1 h, after which Lpcat3 KO enterocytes were found to contain significantly less [ 14 C]cholesterol compared with controls (25%, p Ͻ 0.05; Fig. 6A), indicating a defect in cholesterol uptake. The radiolabeled cells were then chased in the presence of oleic acid for 2 h. Lpcat3 KO enterocyte medium contained much less [ 14 C]cholesterol compared with control medium (45%, p Ͻ 0.01; Fig. 6B). The conditioned medium was then subjected to density gradient ultracentrifugation to determine the effect of Lpcat3 ablation on cholesterol secretion with chylomicrons and HDL. Cholesterol secreted by control enterocytes was distributed in two separate fractions that corresponded to chylomicrons (fractions 1 and 2) and HDL (fractions 8 -10) (Fig. 6C). Similar analyses with Lpcat3 KO enterocytes revealed that cholesterol secretion as part of chylomicron and HDL was significantly reduced (81 and 80% respectively, p Ͻ 0.01, Fig. 6C). These studies indicated that cholesterol uptake and secretion (through the chylomicron and HDL pathways) were diminished in the Lpcat3-deficient animals.
It has been reported that Lpcat3 protects against ER stress in the liver (8). To determine whether this also applies to the small

Lpcat3 Deficiency and Lipid Metabolism
intestine, we measured mRNA levels for three well known ER stress markers, namely PERK, IRE-1␣, and BIP. PERK and IRE-1␣ levels were significantly reduced in Lpcat3 KO enterocytes compared with controls, whereas the BIP level was not significantly altered (Fig. 7).
Lpcat3 Deficiency in the Liver Decreases Plasma Triglyceride but Not Other Lipid Levels-We also injected (intraperitoneally) AAV-Cre (1 ϫ 10 12 ) into the Lpcat3-Flox mice to prepare liver-specific Lpcat3-deficient mice. AAV-LacZ was used as controls. We measured liver and small intestine Lpcat3 mRNA levels in AAV-LacZ and AAV-Cre-treated mice (after 1 month) using real time PCR, and we found the liver-specific Lpcat3 depletion (Fig. 8A). Moreover, we found that AAV-Cre-treated Lpcat3-Flox male animals significantly decreased liver total Lpcat activity (67%, p Ͻ 0.001) and plasma triglyceride (40%, p Ͻ 0.05) but not other lipid and apolipoprotein levels (Fig. 8,  B-D). Also, we did not find any significant changes in liver triglyceride, cholesterol, and phospholipid levels between the two groups (Fig. 8E). We stained the livers and did not find morphological difference between liver Lpcat3-deficient and control mice (Fig. 8F). These results were consistent with a previous report under chow diet (9).

Discussion
PC is the major phospholipid of mammalian cell membranes and also is abundant in circulating lipoproteins (20,21). Although it is not completely understood why so many different PC species exist in nature, it is likely that the relative levels of these different PC species in membranes determine membrane fluidity and thus help cells adapt to environmental changes or other physiological needs. With the cloning of Lpcat family members in recent years, it has become clear that these enzymes are responsible for PC remodeling and the generation of polyunsaturated PCs in various tissues. Lpcat3 is the major Lpcat isoform in the small intestine and liver (6). It has been shown that ablation of Lpcat3 in the small intestine at early stages of life is neonatally lethal and diminishes the incorporation of arachidonic acid into PCs within the cell membrane, leading to the impaired triglyceride mobilization and lipoprotein production (9,10). We previously reported that the neonatal lethality of global Lpcat3 KO mice could be rescued. However, the KO mice had shorter and wider small intestinal villi and a longer and larger small intestine (11). Thus, the impact of Lpcat3 deficiency on small intestine-related lipoprotein metabolism in adult animals remains unknown.
In this study, we used an inducible intestine-specific Lpcat3deficient and a liver-specific Lpcat3-deficient mouse model to investigate the role of PC remodeling on plasma lipoprotein metabolism. Our major findings are as follows. 1) Small intestine-specific Lpcat3 deficient in adult mice have dramatically reduced all plasma lipid and lipoprotein levels, whereas liver deficiency only reduces plasma triglyceride levels at a lower extent compared with intestinal deficiency (40% versus 81%) (Figs. 2D and 8B). A previous report indicated that albumin-Cre-mediated liver Lpcat3 deficiency only caused about 25% reduction of plasma triglyceride levels with no changes of other lipids (9). Thus, manipulation Lpcat3 activity in the small intestine has a dominant effect on plasma lipid metabolism. 2) Under a chow diet-fed condition, Lpcat3 deficiency in adult mice do not cause lipid accumulation in the small intestine and liver.
3) The small intestine deficiency results in a significant reduction of NPC1L1, CD36, ABCA1, and ABCG8 in the enterocyte plasma membrane, thus reducing lipid uptake and reducing cholesterol secretion through both apoB-dependent and apoB-independent pathways. It is conceivable that inhibition of small intestine Lpcat3 might constitute a novel approach for treating hyperlipidemia.
Dietary lipid absorption occurs in the lumen of the small intestine and on the apical surface of enterocytes. NPC1L1 and ABCG5/8 are the two major factors mediating net cholesterol uptake. The former mediates cholesterol uptake, and the latter mediates excretion of excessive cholesterol into the intestinal lumen (22,23). CD36 also participates in cholesterol uptake at the brush border of enterocytes (24,25). Three proteins, namely CD36 (25), plasma membrane-associated fatty acidbinding protein (26), and FATP4 (27), are involved in free fatty acid uptake by enterocytes. The former two proteins are located on the apical surface of enterocytes, and the latter is located on the ER membrane (24). It is conceivable that ablation of Lpcat3 may reduce the incorporation of polyunsaturated PCs in the enterocyte plasma membrane with a consequent increase in membrane rigidity, which may perturb endocytosis and protein recruitment to the plasma membrane and ultimately lead to defective lipid uptake and secretion. Indeed, intestinal cholesterol absorption is mediated by endocytosis of NPC1L1 (28), and change in the composition of polyunsaturated PCs affects the endocytosis process (29,30).
Although the Lpcat3-deficient small intestine did not show obvious morphological alterations (Fig. 1F), the levels of both NPC1L1 and CD36 were significantly reduced in Lpcat3-deficient primary enterocytes compared with control mice (Fig. 4,   FIGURE 7. Small intestine lipid and ER stress marker measurements. Lpcat3-Flox/Villin-Cre-ER T2 and control female mice were treated with tamoxifen. At day 9, small intestines were used to measure ER stress markers. Values are means Ϯ S.D. n ϭ 5. *, p Ͻ 0.05. A and B). Immunofluorescence staining revealed that NPC1L1 abundance was significantly reduced in Lpcat3-deficient villi compared with the control (Fig. 4C). These results represent a plausible explanation for the observed reduction in cholesterol and triglyceride absorption. The reduction of ABCG8 abundance (Fig. 4, A and B) could explain the compensation for defective cholesterol uptake in Lpcat3-deficient enterocytes. It has been shown that global deletion of Lpcat3 at an early stage of development results in the accumulation of lipid droplets in enterocytes (9,10). However, we did not find any difference in lipid levels in the small intestine of Lpcat3-deficient and control mice (Fig. 2G). Thus, the effects of Lpcat3 deletion differ depending on whether the gene is deleted in utero or in adult tissues. This is not an uncommon phenomenon; indeed, depleting liver kinase B1 at different stages of life results in completely different phenotypes (31,32).
Lpcat3 deficiency in the small intestine significantly decreased the concentration of plasma apoA-I (Fig. 2E). It has been demonstrated that the small intestine is one of the sources of apoA-I in blood (33,34). Plasma apoA-I levels in healthy humans increases after ingestion of a fatty meal (35). Moreover, apoA-I can be transferred from chylomicrons to HDL (36). Thus, in our Lpcat3 KO mice, the observed significant reduction in plasma apoA-I level is the major outcome of defective absorption of cholesterol, triglycerides, and phospholipids (Fig.  3, A-D).
Does Lpcat3 deficiency affect lipid secretion? Based on studies of the small intestine in neonatal mice, two groups of researchers proposed models for the assembly of Lpcat3-related apoB-containing lipoproteins (9,10). However, cholesterol secretion in cultured primary enterocytes is mediated by both apoB-dependent and -independent pathways (14). In this study, we found that Lpcat3 deficiency in primary enterocytes reduced not only cholesterol uptake (Fig. 6A) but also secretion (Fig. 6B). Moreover, the deficiency reduced cholesterol secretion through both the chylomicron and HDL pathways (Fig.  6C). This reflects the fact that Lpcat3 deficiency, at least in part, reduces polyunsaturated PCs within the enterocyte plasma membrane (Table 1). This modification influences basal as well as apical membrane PC composition. Indeed, we found a significant reduction in the level of ABCA1 (Fig. 4, A and B), which is located on the basal membrane of enterocytes and is a major player in the apoB-independent pathway (14,37).
The results obtained from our liver Lpcat3-deficient mice (Fig. 8, A-D) echoed what has been reported previously (9), showing a reduction in total liver Lpcat activity and plasma triglyceride but no other lipid levels. Importantly, we also confirmed that Lpcat3 deficiency in the liver does not cause signif-

Lpcat3 Deficiency and Lipid Metabolism
icant lipid accumulation in the liver under chow diet conditions (Fig. 8E) (9). Why then does liver deficiency have a smaller effect than that of the small intestine in terms of plasma lipid metabolism? One explanation could be due to the compensation of other Lpcats. Indeed, we previously reported that global Lpcat3 deficiency causes about 90% reduction of total Lpcat activity in the small intestine, whereas the deficiency only causes about a 70% reduction of total activity in the liver (11). We confirmed this phenomenon in this study (Fig. 1E and 8B).
Acute knockdown of liver Lpcat3 in genetically obese mice exacerbates lipid-induced ER stress (8). However, the same group recently reported that Lpcat3 deficiency in the liver does not affect the expression of ER stress markers (9). We also recently reported that global Lpcat3 deficiency has no impact on ER stress (11). In this study, we again measured the expression of ER stress markers in the Lpcat3-deficient small intestine and found that the deficiency attenuated (rather than enhanced) the expression of PERK and IRE-1␣ (Fig. 7).
In this study, the Lpcat3 depletion was achieved by tamoxifen treatment in adult Lpcat3-Flox/Villin-Cre-ER T2 mice over a period of 9 days so that the effect was acute, similar to drug intervention. Given the fact that small intestine Lpcat3 deficiency has 1) a much bigger impact on plasma lipid levels than that of liver deficiency, 2) no effect on lipid retention in the tissue, and 3) a marginal effect on ER stress, selective inhibition of Lpcat3 activity in the small intestine could be a novel approach for treating hyperlipidemia.
Author Contributions-I. K. and Z. L. designed and performed experiments, analyzed data, interpreted results, discussed implications, and critically evaluated the manuscript. H. H. B. and M. S. K. used measured PC subspecies using LC/MS/MS and reviewed the manuscript. G. G. provided AAV-LacZ and AAV-Cre and reviewed the manuscript. X. C. J. conceived the study, supervised the project, interpreted results, discussed implications, and wrote the manuscript.