Selective evaluation of high density lipoprotein from mouse small intestine by an in situ perfusion technique.

The small intestine (SI) is the second-greatest source of HDL in mice. However, the selective evaluation of SI-derived HDL (SI-HDL) has been difficult because even the origin of HDL obtained in vivo from the intestinal lymph duct of anesthetized rodents is doubtful. To shed light on this question, we have developed a novel in situ perfusion technique using surgically isolated mouse SI, with which the possible filtration of plasma HDL into the SI lymph duct can be prevented. With the developed method, we studied the characteristics of and mechanism for the production and regulation of SI-HDL. Nascent HDL particles were detected in SI lymph perfusates in WT mice, but not in ABCA1 KO mice. SI-HDL had a high protein content and was smaller than plasma HDL. SI-HDL was rich in TG and apo AIV compared with HDL in liver perfusates. SI-HDL was increased by high-fat diets and reduced in apo E KO mice. In conclusion, with our in situ perfusion model that enables the selective evaluation of SI-HDL, we demonstrated that ABCA1 plays an important role in intestinal HDL production, and SI-HDL is small, dense, rich in apo AIV, and regulated by nutritional and genetic factors.


In situ perfusion system for the evaluation of HDL production in the SI
We developed an in situ perfusion model in mice with isolated SI in which the arterial blood supply for the SI has been blocked for the assessment of HDL production in mouse SI ( Fig. 1A and supplementary Fig. I). This novel in situ perfusion model can be used for the evaluation of HDL originated from the SI without intrusion by the fi ltration of HDL from plasma. The detailed experimental procedures are available in the form of video upon request. After the mice were deeply anesthetized, the abdominal cavity was opened, and the appropriate arteries except for the abdominal aorta were ligated to block the blood supply for the SI ( Fig. 1A ). Argatroban was injected intravenously as an anticoagulant ( 23 ), and 5 min after injection, the mice were euthanized for the removal of tissues including the heart and lungs from the thorax, except for the thoracic descending aorta. At this point, the SI would become ischemic.
To protect the SI from organ damage, the following procedures needed to be performed quickly. Two tubes were inserted into the abdominal aorta and portal vein by the following procedure. First, a needle (26G) connected to a tube was inserted antegrade through the thoracic descending aorta into the abdominal aorta, which was then ligated with the needle inserted at distal to the inserted portion to secure the needle as the inlet for perfusion (supplementary The intestinal lymph duct was then immediately cannulated to serve as the outlet with a needle (26G) connected to a silicon tube in the same manner as for the portal vein (supplementary Fig. I), while a buffer solution (DMEM without phenol red containing 0.03% BSA) aerated with 95% O 2 and 5% CO 2 was perfused slowly from the inlet. The interval when the mice were euthanized to circulation of the infusion had to be less than 5 min.
After all three tubes were secured, perfusion was started, and the fl ow rate of the perfusion buffer was increased to 0.5 ml/min. The body was kept at 35°C in Krebs Ringer buffer in a water bath, and sample collection by means of gravity was started from the two outlets. We confi rmed that the motility of the intestines, glucose uptake, and an appropriate ratio of lymph to portal perfusates (1:50) could be maintained for up to 90 min in our system. However, due to the fragility and sensitivity of the SI ( 24,25 ), prolonged perfusion seems to be problematic, and the SI showed organ damage such as edema and lower motility 90 min after perfusion. Therefore, we collected samples from 10 min through 60 min after the start of perfusion. The collected samples were pooled for the analyses.
The production of HDL from the SI was examined by analyzing HDL-apo AI and apo AI using non-SDS-PAGE and SDS-PAGE, respectively, followed by Western blot analysis of apo AI in SI lymph perfusates from WT mice and ABCA1 KO mice. Peptide mapping of HDL using LC/MS and the analyses of lipid and apo composition and size distribution of HDL were performed in plasma, SI lymph perfusates, and liver perfusates from WT mice. The effects of an ABCA1 inhibitor, glyburide ( 26 ) (Cayman), and an LCAT inhibitor, DTNB ( 27 ) (Cayman), on the formation of SI-HDL were examined in WT mice by adding them to the perfusion buffer. The effects of a high-fat diet on the production of SI-HDL were examined by measuring HDL-apo AI in SI lymph perfusates using non-SDS-PAGE followed by Western blot analysis of apo AI after WT mice were fed a high-fat diet for 4 weeks.

Western blot analyses
For the analysis of apos in SI lymph perfusates, plasma, and lipoprotein density fractions, samples were run on SDS-PAGE or prerequisite molecules for the production of HDL (6)(7)(8). We and others found that a human intestinal cell line, CaCo-2, expresses ABCA1 ( 9,10 ) and secretes HDL though the involvement of ABCA1 ( 11 ). Experimental studies with rodents have demonstrated that the liver and small intestine (SI) are the major organs for ABCA1 expression and HDL production (12)(13)(14)(15)(16). Recent fi ndings have shown that, although the overexpression of hepatic ABCA1 in experimental mouse models is not benefi cial ( 17 ), the overexpression of intestinal ABCA1 is benefi cial for atherosclerosis ( 18 ), indicating that HDL from SI may perform unique and distinct functions.
However, the characterization of SI-derived HDL (SI-HDL) and clarifi cation of the mechanism that underlies its production and regulation remain elusive because selective evaluation of SI-HDL is diffi cult. Since the 1970s, various studies have demonstrated that HDL can be obtained from the intestinal lymph duct of anesthetized rodents in vivo, but its origin has been doubtful because it has been suggested that the source is either the secretion of HDL by the SI or the fi ltration of HDL from plasma through the blood capillary-lymph loop into the intestinal lymph duct (19)(20)(21)(22).
In this regard, the group of Hayden et al. ( 16 ) recently reported the interesting and unexpected fi nding that HDL from the intestinal lymph duct was not reduced in SI-specifi c ABCA1 KO mice but was markedly reduced in liver-specifi c ABCA1 KO mice. The authors concluded that SI-HDL may be secreted directly into the circulation and that HDL in the intestinal lymph duct is predominantly derived from plasma. However, this conclusion is tenable only if a substantial quantity of plasma HDL, the majority of which is derived from the liver, passes through the loop and enters the intestinal lymph duct, so that the liver-specifi c disruption of ABCA1 affects the quantity of HDL obtained from the intestinal lymph duct more than does the SI-specifi c disruption of ABCA1.
To prevent the possible fi ltration of plasma HDL into the intestinal lymph duct and to realize the selective evaluation of SI-HDL, we developed an in situ per fusion technique using surgically isolated mouse SI. With this technique, we could demonstrate that the SI produces HDL, which reaches at least the SI lymph duct, and that the production of SI-HDL may be dynamically regulated.

Animals
WT C57BL6/J mice, ABCA1 KO, and apo E KO mice were obtained from the Jackson Laboratory (Bar Harbor, ME) and housed in a room under controlled temperature and humidity conditions and with free access to water and chow. Experiments were conducted when the male mice were aged from 12 to 16 weeks. Blood was drawn for the measurement of plasma lipid levels by enzymatic methods. The experimental protocol was approved by the Ethics Review Committee for Animal Experi mentation of Osaka University.
swelling of the liver, the abdominal vena cava was incised immediately after cannulation, the thorax was opened, a part of the inferior vena cava was cut, and a cannula was inserted.
After the collected perfusate was concentrated using centrifugal fi lter devices (Millipore), HDL was separated by ultracentrifugation and used for the analysis of HDL proteins using LC/MS, analysis of the HDL lipid composition using HPLC, and analysis of the HDL apo composition.

MS analysis of HDL fractions
HDL fractions from mouse serum, SI lymph perfusate, SI lymph ( 14 ), and liver perfusate ( 28 ) from WT mice were buffer exchanged with 10 mM triethylammonium bicarbonate buffer (Sigma-Aldrich, St. Louis, MO) and concentrated by centrifugation (6,000 g , 30 min at 4°C) using Vivaspin2 (MW 3000; GE Healthcare, Buckinghamshire, UK). The protein concentration of each sample was determined with the Bradford method and adjusted to 7.5 mg/ml. Ten microliters of each sample was then subjected to reduction and alkylation [1 l of denaturant and 2 l of reducing reagent (AB Sciex, Foster City, CA)] for 60 min at 60°C, cystein blocking [1 l of cystein-blocking reagent (AB Sciex) ], and trypsinization [1.9 g of trypsin (Roche, Basel, Switzerland)] for 4 h at 37°C followed by the addition of another 1.9 g of trypsin and overnight incubation at 37°C. After digestion, non-SDS-PAGE, transferred to polyvinylidene difl uoride membranes, and immunoblotted with anti-mouse apo AI, apo AIV, apo E, apo B100, or apo B48 overnight at 4°C. Membranes were washed and then incubated with anti-IgG antibody conjugated with HRP for 1 h at room temperature. An ECL Advance Western Blotting Detection Kit (Amersham Biosciences, Piscataway, NJ) was used for the visualization of immunoblots according to the manufacturer's protocol.

Conventional SI lymph cannulation experiments
After intraperitoneal anesthesia, the main mesenteric lymphatic duct of WT mice was cannulated as previously described by Green et al. ( 14 ). HDL was separated by ultracentrifugation from a sample of mesenteric lymph and used for the analysis of HDL proteins using LC/MS.

In situ perfusion system for the evaluation of HDL production in the liver
In situ perfusion for the liver from WT mice was performed according to the method reported by Sugano et al. ( 28 ) with minor modifi cations. Briefl y, after the mice were deeply anesthetized, the abdomen was opened, the portal vein was cannulated in situ, and the liver was perfused with oxygenated buffer, which was the same as that used for in situ perfusion of the SI. To prevent microscopy (EM), as described previously ( 35 ). In brief, for transmission electron microscopy (TEM), HDL was separated by preparative ultracentrifugation and dialyzed against saline contain ing 1 mM EDTA (pH 8.0) overnight at 4°C to remove KBr. Next, HDL was dialyzed against a 10 mM NH 4 HCO 3 solution for 2 h at 4°C and negatively stained with 1% uranium acetate. Electron micrographs were obtained with a computer-controlled JEOL 1200EX electron microscope (JEOL Inc., Tokyo, Japan). Images at a fi nal magnifi cation of 200,000× were acquired with a highresolution digital camera. The diameters of spherical HDL particles were measured using TEM imaging Platform iTEM (Olympus Soft Imaging Solutions GmbH, Münster, Germany).

Two-dimensional gel electrophoresis
To examine the effects of LCAT inhibition on SI-HDL production, HDL in SI lymph perfusates collected using the in situ perfusion technique from WT mice with and without the presence of DTNB in the perfusion buffer was separated by native twodimensional gel electrophoresis as described previously ( 41,42 ). Fresh SI lymph perfusates were run on an agarose gel (0.75%) and then on a 2% to 25% polyacrylamide gel at 0°C at 100 V for 20 h. Fractionated HDL was electroblotted to a nitrocellulose sheet at 0°C and detected with the following antibodies. The fi rst antibody was a rabbit anti-mouse apo AI antibody (BioDesign), and the second was a goat anti-rabbit IgG (DAKO) iodinated with Na 125 I by a modifi ed chloramine T method.

Statistical data analysis
Statistical data analyses were performed using the Statistical Analysis System (SAS) Software Package (Ver. 9.2; SAS Institute Inc., Cary, NC) at Fukuoka University (Fukuoka, Japan). Differences in the lipid and protein composition of HDL between groups were examined by an ANOVA using the general linear model ( 43 ). Differences in the size of HDL particles between plasma HDL and SI-HDL were examined by the Wilcoxon rank sum test. Data are presented as the mean ± SD, and the signifi cance level was considered to be <0.05 unless indicated otherwise.

Development of a novel in situ perfusion model in mice
A novel in situ perfusion model was developed in mice with isolated SI in which the arterial blood supply for the SI is blocked, leaving only the superior mesenteric artery open as the perfusion inlet for the SI. Fig. 1A shows a schematic representation of our novel in situ perfusion model. Because the source of HDL obtained from the mesenteric lymph duct of anesthetized rodents in vivo is either the secretion of HDL by SI or the fi ltration of HDL from plasma through the blood capillary-lymph loop into the intestinal lymph duct (19)(20)(21)(22), in our in situ perfusion model the arterial blood supply for the SI was blocked by ligation of the abdominal aorta and its branches ( Fig. 1A ) to dissociate the HDL produced by the SI from HDL fi ltered from plasma.
The superior mesenteric artery was not ligated and was left open as the perfusion inlet ( Fig. 1A ). A 26G needle was inserted antegrade from the thoracic descending aorta into the abdominal aorta before ligation of the abdominal aorta and used to pump perfusion buffer through the the samples were desalted and concentrated with C18 Stage tips ( 29 ) packed in-house in 60 l of 2% acetonitrile (ACN) and 0.1% trifl uoroacetic acid (TFA) buffer.

Lipid profi le analyses using HPLC
Two hundred microliters of SI lymph perfusates or HDL separated by ultracentrifugation from plasma, SI lymph perfusates, and liver perfusates from WT mice was analyzed with the HPLC system using two tandem gel permeation columns (Lipopropak XL, 7.8 mm × 300 mm; Tosoh Corp., Tokyo, Japan) at a fl ow rate of 700 l/min. Total cholesterol (TC), TG, free cholesterol (FC), and phospholipid (PL) were measured with two parallel online enzymatic lipid detection systems (350 l/min each) (Skylight Biotech, Inc., Akita, Japan) ( 31-34 ). The system was calibrated with the aid of latex beads and high-molecular-weight standards for the apparent spherical diameters of HDL.

Isolation of lipoproteins by ultracentrifugation
Lipoprotein fractions were isolated from SI lymph perfusates, SI lymph, liver perfusates, and plasma from WT mice by serial preparative ultracentrifugation as described previously (35)(36)(37)(38)(39). Briefl y, SI lymph perfusate was overlaid with saline solution at a volume ratio of ‫ف‬ 5:3 and ultracentrifuged at 50,000 g for 25 min (2.25 × 10 6 g/min) ( 40 ) at 10°C in a TLA-100.2 rotor in a Beckman TL-100 Tabletop Ultracentrifuge (Beckman Instruments Inc. ). Chylomicron (CM) fraction (upper fraction) was collected using a tube slicer. The bottom fraction was overlaid with saline and ultracentrifuged at 100,000 rpm for 2 h at 10°C. The VLDL fraction (upper fraction) was collected using a tube slicer. The density of the d > 1.006 g/ml fraction (bottom fraction) was then adjusted to 1.063 g/ml with solid KBr and overlaid with d = 1.063 g/ml KBr solution. After centrifugation at 100,000 rpm for 2 h at 10°C, the LDL fraction (upper fraction) was collected. Finally, the density of the d >1.063 g/ml fraction (bottom fraction) was adjusted to 1.25 g/ml with solid KBr and overlaid with d = 1.25 g/ml KBr solution. After centrifugation at 100,000 rpm for 5 h at 10°C, the HDL fraction (upper fraction) was collected with a tube slicer. LDL and HDL fractions were dialyzed against saline containing EDTA (1 mM) to eliminate KBr. HDL for lipid profi le analyses using HPLC and determination of protein concentration was used without dialysis.

Electron microscopy of HDL particles
The size distributions of HDL particles separated from SI lymph perfusates and plasma of WT mice were examined by electron Therefore, we used ABCA1 KO mouse bodies for in situ perfusion but added serum from WT mice to the perfusion buffer. Perfusates from the abdominal aorta (inlet) and lymph duct and portal vein (two outlets) were run on nondenaturing PAGE followed by Western blot analysis for apo AI and on SDS-PAGE followed by Western blot analysis for apo B100 and apo B48 ( Fig. 1C ). As shown in the upper panel of Fig. 1C , a substantial amount of HDL was present in the perfusates collected from both the SI lymph duct and the portal vein. Because SI-HDL is not formed in ABCA1 KO mice due to the defect in the lipidation of apo AI, HDL detected in SI lymph perfusates should come from the infi ltration of plasma HDL from the abdominal aorta.
Apo B was detected in perfusates collected from the portal vein, but not in SI lymph perfusates ( Fig. 1C , lower panel). Because no apo B was detected in SI lymph perfusates, our result indicates that apo-B-containing lipoprotein was not fi ltrated from the abdominal aorta into the lymph duct.

Mapping of peptides from intestinal HDL and hepatic HDL using LC/MS
A previous study has shown that HDL from the intestinal lymph duct obtained in vivo from anesthetized mice is likely to contain HDL from the systemic circulation, most of which was derived from the liver ( 16 ). Therefore, we compared the SI-HDL obtained using our novel in situ perfusion mouse model, the intestinal HDL (C-HDL ) obtained in vivo from anesthetized mice using a conventional experimental procedure, and hepatic HDL (L-HDL) obtained from liver perfusion. Protein moieties of HDL from mouse plasma (P-HDL), SI-HDL, C-HDL, and L-HDL were compared by using LC/MS ( Fig. 2 ).
As shown in Fig. 2 , the peptide patterns of C-HDL were very similar to those of L-HDL: the same number of major peptides was detected ( m/z 543, 402, 403, 523, 413, 422, and 435), and they had similar relative peptide-ion intensities. These results suggest that C-HDL may contain HDL from the systemic circulation, most of which is derived from the liver ( 16 ).
However, the peptide patterns of SI-HDL were apparently different from those of C-HDL and L-HDL: SI-HDL had additional peptides of m/z 542 and 524 that were not detected in C-HDL and L-HDL ( Fig. 2B , indicated by red arrows). Because HDL obtained using our novel in situ perfusion technique is not subject to interference from the liver and plasma, our results indicate that intestinal HDL is different from hepatic HDL and that the novel in situ perfusion model is suitable for the selective evaluation for SI-HDL.

Distribution of lipids and apos in lipoproteins produced from the SI
To characterize lipoproteins produced from the SI, we examined the distribution of lipids and apos in lymph perfusates collected from WT mice using our novel in situ perfusion model. Lipid profi les were analyzed by online monitoring for TC, FC, TG, and PL after lipoproteins abdominal aorta into the mesenteric artery ( Fig. 1A ). The portal vein and intestinal lymph duct were punctured and cannulated to serve as outlets ( Fig. 1A ). Therefore, in our in situ perfusion mouse model, no further systemic blood will come into the SI after perfusion with a buffer solution starts, and thus the HDL in the infusates collected from the SI lymph duct (one of the outlets) would only be produced from the SI and not infi ltrate from the systemic plasma.

The SI produces HDL
The intestine and liver are the two major sites of HDL production ( 15,16,44 ). To demonstrate that the SI produces HDL, SI lymph perfusates were collected from WT mice using our in situ perfusion model. Nondenaturing PAGE followed by Western blot analysis for apo AI, the major protein of HDL, clearly showed that HDL from the SI was present in SI lymph perfusate ( Fig. 1B , left lane of the upper panel), but not in perfusate collected from the portal vein (data not shown). Because systemic blood does not come into the SI during perfusion due to the blockade of systemic blood by ligation of the abdominal aorta and its branches in our in situ perfusion model, this result indicates that the SI produces HDL in WT mice.

ABCA1 is required for the production of HDL from the SI
Intestinal ABCA1 has been shown to contribute to HDL biogenesis in mice in vivo ( 16 ). To clarify the infl uence of ABCA1 on the production of HDL from the SI, we collected lymph perfusates from ABCA1 KO mice using our in situ perfusion model. HDL-apo AI was not detected in SI lymph perfusates from ABCA1 KO mice ( Fig. 1B , right lane of the upper panel). This result indicates that ABCA1 is required for the production of HDL from the SI.
However, an apo AI immunoreactive mass was detected in the form of doublets in the SI lymph perfusate from ABCA1 KO mice, in contrast to a major single band that was detected in that from WT mice ( Fig. 1B , lower panel). It is possible that the upper band in ABCA1 KO mice may represent a precursor of apo AI, newly synthesized from the intestine. Therefore, the defi cient HDL production from the SI of ABCA1 KO mice was not due to the defective synthesis of apo AI, but rather to the defective lipidation of apo AI ( 45 ).

Evidence that plasma HDL can fi ltrate from the abdominal aorta into the SI lymph duct
Although a previous study suggested that plasma HDL can infi ltrate into the mesenteric lymph ( 16 ), there is still some controversy, and direct evidence is not available. We used our in situ perfusion model to clarify whether plasma lipoproteins in abdominal aorta contribute to HDL in SI lymph. Because ABCA1 KO mice showed the defi cient production of HDL from the SI ( Fig. 1B ), the addition of serum from WT mice to the perfusion buffer should be able to show whether plasma HDL can infi ltrate from the abdominal aorta to the mesenteric lymph duct.
the SI contained only apo B48, and no apo B100, as expected ( 46 ). Apos were not detected in the LDL fraction separated from lymph perfusates ( Fig. 3B ), indicating that the LDL-size fraction was not produced by the SI. HDL from SI lymph perfusates contained both apo AI and AIV, similar to plasma HDL ( Fig. 3B ).

Lipid and protein composition of SI-HDL
WT mice were used to examine the lipid and protein composition of HDL produced from the SI. HDL was separated from plasma, SI lymph perfusates collected using our in situ perfusion model, and liver perfusates using small-scale preparative ultracentrifugation. HDL separated by ultracentrifugation was used for the measurement of total protein and apos but was further separated using HPLC for the online were separated by HPLC ( Fig. 3A ). As shown in Fig. 3A , lipoproteins in SI lymph perfusates were separated into two main fractions, one corresponding to plasma HDL and another corresponding to plasma non-HDL (CM to VLDL size). The four main classes of lipids (i.e., cholesterol ester [CE], FC, TG, and PL) were mainly distributed in the non-HDL fraction ( Fig. 3A ), similar to plasma lipoproteins.
To examine the distribution of apos, lipoprotein density fractions (CM, VLDL, LDL, and HDL) were separated from lymph perfusates of WT mice by small-scale preparative ultracentrifugation. Apo AI, apo AIV, apo B100, and apo B48 in each lipoprotein density fraction were detected by Western blot analysis after separation by SDS-PAGE ( Fig. 3B ). As shown in Fig. 3B , lipoproteins produced from protein in SI-HDL and L-HDL were signifi cantly lower than those in P-LDL ( Table 1 ). Table 1 also shows the distribution of lipids in HDL. At ad libitum, SI-HDL tended to show less CE and more TG than P-HDL and L-HDL ( Table 1 ). The distribution of CE and TG was signifi cantly ( P < 0.05) different between SI-HDL and L-HDL at ad libitum and between SI-HDL at ad libitum and at fasting, as assessed by a two-way ANOVA (data not shown). At fasting, TG was not detected in L-HDL but was detected in SI-HDL ( Table 1 ).
Therefore, using our in situ perfusion model, we showed that SI-HDL was protein rich compared with HDL in plasma in WT mice and TG rich compared with L-HDL. measurement of lipids to ensure that lipids in HDL are measured without possible interference from other lipoproteins. Supplementary Fig. II gives examples of the HPLC TC profi le of HDL separated from plasma, lymph perfusates, and liver perfusates to show the purity of HDL. HPLC-separated HDL was measured for TC, TG, FC, and PL, and CE was calculated from TC and FC ( 47 ). Table 1 and Fig. 4A show the contents of lipids and total protein in P-HDL, L-HDL, and SI-HDL. SI-HDL, similar to L-HDL, had a signifi cantly higher protein content and lower lipid content than P-HDL ( Table 1 , Fig. 4A ). CE and PL were the major lipids in SI-HDL and L-HDL, similar to P-HDL. However, the contents of CE and PL relative to  3. HPLC analyses of lipids and preparative ultracentrifugation followed by Western blotting for apos in SI lymph perfusates. A: HPLC analyses of SI lymph perfusates from WT mice. Two hundred microliters of SI lymph perfusates was run on HPLC as described in the Methods. TC, FC, TG, and PL were measured enzymatically. Arrows denote the average elution time of indicated plasma lipoproteins in WT mice. B: Apo distribution among lipoproteins in SI lymph perfusates from WT mice separated by preparative ultracentrifugation. Lipoproteins in plasma and SI lymph perfusates were subjected to small-scale preparative ultracentrifugation to concentrate samples, and the concentrated samples were then run on SDS-PAGE followed by Western blot analysis using antibodies against the indicated apos. CM, VLDL, LDL, and HDL denote the density range of the indicated plasma lipoprotein fractions. compared with that of P-HDL, and the average size of SI-HDL particles (mean ± SD: 11.06 ± 2.70 nm) was significantly ( P < 0.001) smaller than that of P-HDL particles (12.94 ± 1.64 nm). This result indicates that the particle size of SI-HDL was smaller than that of P-HDL.

Inhibitors of ABCA1 and LCAT affect the formation of SI-HDL
Because we have shown that most of the HDL particles secreted from the SI are spherical using EM ( Fig. 5A ), to identify the mechanism for SI-HDL assembly, we examined the effects of inhibitors of ABCA1 and LCAT on the formation of SI-HDL. It is well known that ABCA1 lipidates apo AI to form HDL and LCAT converts lipid-poor pre-␤migrating HDL to mature ␣ -migrating HDL. We used glyburide ( 26 ) and DTNB ( 27 ) as inhibitors of ABCA1 and LCAT, respectively ( Fig. 6A ). The effects of ABCA1 and LCAT inhibitors were examined by collecting SI lymph perfusates from WT mice that underwent in situ perfusion using buffers with and without the presence of the inhibitors ( Fig. 6 ).
As shown in Fig. 6B , nondenaturing PAGE followed by Western blot analysis for apo AI showed that there was a marked increase in free apo AI and small HDL in SI lymph perfusates in the presence of glyburide compared with the absence of glyburide. This result suggests that ABCA1 is involved in the lipidation of apo AI to form SI-HDL.
As shown in Fig. 6C , two-dimensional electrophoresis of SI lymph perfusates followed by Western blotting for apo AI showed that the presence of premature HDL such as pre-␤ 1-and pre-␤ 2-HDL ( 41,42 ) was observed in the presence of DTNB, but not in the absence of DTNB. A reduction in ␣ -HDL particle size was also observed in the

Apo composition of SI-HDL
To characterize the apo composition of SI-HDL, L-HDL and SI-HDL separated by ultracentrifugation from liver perfusates and SI lymph perfusates, respectively, were run on SDS-PAGE and then subjected to Western blot analysis for apo AI, apo AIV, and apo E ( Fig. 4B ). As shown in Fig. 4B , L-HDL contained a very limited amount of apo AIV but a considerable amount of apo E, whereas an opposite trend was seen for SI-HDL. This result indicates that SI-HDL was rich in apo AIV compared with L-HDL, and SI-HDL is different from L-HDL with respect to the composition of apos.

Size distribution of SI-HDL
Because SI-HDL was protein rich compared with P-HDL ( Table 1 ), we used EM to examine the size distribution of HDL separated using ultracentrifugation from SI lymph perfusates and plasma of WT mice ( Fig. 5 ). Fig. 5A shows the representative negative-stain electron micrographs of SI-HDL and P-HDL. As shown, SI-HDL particles were a population of spheres with a very small number of discs. We measured the particle diameter of spherical particles in SI-HDL and P-HDL ( Fig. 5B, C ). Fig. 5B shows the size distribution of SI-HDL and P-HDL. As shown, SI-HDL particles, similar to P-HDL, were heterogeneous in size, but the distribution of SI-HDL particles was more diverse than that of P-HDL, and the size distributions of SI-HDL and P-HDL overlaid ( Fig. 5B ). However, SI-HDL apparently had a higher proportion of smaller particles as compared with P-HDL ( Fig. 5B ). Fig. 5C shows the individual data and the box plots of SI-HDL and P-HDL. As shown, although the size range of SI-HDL covered that of P-HDL, the peak diameter of SI-HDL particles shifted toward smaller particles as Western blot analysis for apo AI. As shown in Fig. 7A , plasma HDL-C levels and the contents of HDL-apo AI in SI lymph perfusates from WT mice ad libitum were markedly reduced after 24 h fasting. In contrast, a 4-week high-fat diet markedly increased plasma HDL-C levels and the contents of HDL-apo AI in lymph perfusates from WT mice ( Fig. 7B ). These results indicate that fasting reduces and high-fat diet increases the production of HDL from the SI.

Apo E KO reduces the production of HDL from the SI
Apo E KO mice are characterized by a marked reduction of HDL-C levels in plasma. Because intestinal HDL has been shown to signifi cantly contribute to plasma HDL ( 16 ), we used our in situ perfusion model to examine whether presence of DTNB ( Fig. 6C ). This result suggests that LCAT may be involved in the maturation of SI-HDL.
Although our experiments were limited in that the inhibitory effects of ABCA1 and LCAT are unknown, our results suggest that ABCA1 and LCAT may play important parts in the formation of SI-HDL.

Nutritional regulation of the production of HDL from SI
A previous study showed that intestine signifi cantly contributes to plasma HDL-C levels ( 44 ). We examined the effects of fasting and high-fat feeding on the production of HDL from SI in WT mice using our in situ perfusion model to clarify the nutritional regulation of SI-HDL. HDL in lymph perfusates collected under different nutritional conditions was analyzed by non-SDS-PAGE followed by scale bar: 100 nm. B: Size distribution of SI-HDL and P-HDL particles from negative-stain electron micrographs. The frequency distributions of the size of SI-HDL (pink bars; n = 913) and P-HDL (gray bars; n = 1,412) were plotted together, and the red bars represent the overlaid parts. Two measurements were made for the diameter of each HDL particle, and the mean diameter was used to calculate the size frequency. C: Box-and-whisker plots showing the mean ( ), median (middle bar in the rectangle), and 10th (bottom bar), 25th (bottom of rectangle), 75th (top of rectangle), and 90th (top bar) percentiles of the sizes of SI-HDL (black) and P-HDL (red) particles. The individual data are shown on the left of the boxes. * P < 0.001, SI-HDL versus P-HDL, assessed by the Wilcoxon rank sum test.
Our novel in situ perfusion model achieves the selective evaluation of HDL by dissociating HDL production by the SI from the fi ltration of HDL from plasma. In this model, arterial blood supply for the SI is blocked by ligation of abdominal aorta and other arteries, leaving only the superior mesenteric artery open as the perfusion inlet ( Fig. 1A ). Perfusion buffer is pumped through a needle that is connected to a tube and inserted antegrade through the thoracic descending aorta into the abdominal aorta just before ligation of the abdominal aorta ( Fig. 1A ). Therefore, after perfusion starts, no additional systemic blood will enter the SI, and the possible fi ltration of plasma HDL from the systemic circulation into the SI lymph duct can be prevented. The SI lymph duct and portal vein are cannulated as outlets for perfusion buffer ( Fig. 1A ). Under these conditions, the HDL in the infusates collected from the SI lymph duct would originate only from the SI.
Using our in situ perfusion model, we found that HDL was detected in SI lymph perfusates from WT mice ( Fig. 1B ), indicating that the SI produces HDL. This fi nding supports the notion that the intestine, along with the liver, is an important site for the secretion of apo AI and the production of HDL ( 12-16 ). We did not detect HDL in SI lymph perfusates from ABCA1 mice ( Fig. 1B ), indicating that ABCA1 is essential for the production of HDL by the the production of HDL from the SI is reduced in apo E KO mice. As shown in Fig. 7C , apo E KO mice had markedly lower levels of HDL-C and reduced contents of HDLapo AI in SI lymph perfusates as compared with WT mice. These results indicate that apo E may play a role in the biogenesis of SI-HDL.

DISCUSSION
To selectively evaluate HDL produced from the intestine, we developed an in situ perfusion model using surgically isolated mouse SI. Using our in situ perfusion model, we found that the SI produces HDL in mice and ABCA1 plays an important role in the production of SI-HDL, that SI-HDL is different from HDL produced by the liver, and that SI-HDL may be regulated by nutritional and genetic factors.
Our in situ perfusion model using surgically isolated mouse SI was developed for the selective evaluation of SI-HDL because HDL in mesenteric lymph collected from anesthetized mice originates either from the secretion by the SI or from the fi ltration from plasma through the blood capillary-lymph loop into the intestinal lymph duct (19)(20)(21)(22). consider the conventional cannulation experiment to be useful for the analysis of apo-B-containing lipoproteins because we used it to demonstrate an increase in the production of CMs in CD36 KO mice ( 51 ). However, for the selective evaluation of SI-HDL, the newly developed perfusion technique may be the best method available for eliminating the interference of plasma HDL.
Our fi nding that HDL was produced by the SI ( Fig. 1A ) but plasma HDL can also fi ltrate into the SI lymph duct ( Fig. 1B ) resolves controversies regarding intestinalderived HDL (19)(20)(21)(22). A previous study showed that HDL from the intestinal lymph duct obtained in vivo from anesthetized mice is likely to contain HDL from the systemic circulation, the majority of which is derived from the liver ( 16 ). Therefore, peptide mapping of HDL using LC/MS was performed to compare SI-HDL in SI lymph perfusates collected using our in situ perfusion model, C-HDL collected from the SI lymph duct in anesthetized mice, and L-HDL collected from liver perfusates ( Fig. 2 ). C-HDL and L-HDL were very similar in that they had the same number of major peptides and relative peptide-ion intensities ( Fig. 2C, D ). This result agrees with the fi nding of Brunham et al. ( 16 ) that intestinal HDL in mice that lacked intestinal ABCA1 predominantly originates from the liver.
However, we found that SI-HDL in SI lymph perfusates was different from C-HDL and L-HDL in that SI-HDL contained additional major peptides ( m/z 542 and m/z 524) that were not detected in L-HDL or C-HDL ( Fig. 2 ). It is possible that the SI may secrete some known or unknown apos that are not secreted by the liver, and this proposition will need to be examined in future studies.
Our result that the two peptides ( m/z 542 and m/z 524), which were detected as major peaks in SI-HDL obtained using our in situ perfusion model, were not detected in C-HDL obtained from the SI lymph duct in anesthetized mice ( Fig. 2 ), suggests that rate of the production of HDL by the SI is slow as compared with that of fi ltration of preexisting liver-originated HDL from the abdominal aorta into the SI lymph duct, and thus liver-originated HDL is SI. This fi nding supports those of Brunham et al. ( 16 ) that mice that specifi cally lack ABCA1 in the intestine had ‫ف‬ 30% lower plasma HDL-C levels. However, we detected free apo AI in SI lymph perfusates from ABCA1 mice ( Fig. 1B ), indicating that cellular lipids are not available for the lipidation of apo AI to form SI-HDL in the absence of ABCA1 ( 45 ).
Because SI-HDL is not formed in the absence of ABCA1, we used ABCA1 KO mice to clarify whether plasma HDL can fi ltrate from the abdominal aorta through the SI into a lymph duct. We perfused the SI of ABCA1 KO mice using perfusion buffer to which had been added serum from WT mice and found that the collected SI lymph perfusates contained a substantial amount of HDL ( Fig. 1C ). It is possible that lipids in lipoproteins in the aortic perfusate may be delivered to the SI through transintestinal transport, that is, via the transintestinal cholesterol effl ux pathway ( 48 ), and lipidate apo AI from intestinal secretion to form HDL. However, in the absence of ABCA1, which is located at the basolateral membrane of enterocytes ( 49 ), cellular lipids will not be available to free apo AI to form HDL. Therefore, HDL that was detected in SI lymph perfusates of ABCA1 mice perfused with serum from WT mice ( Fig. 1C ) can originate only from fi ltration of plasma HDL. Using our in situ perfusion model, we provided direct evidence that systemic plasma HDL can fi ltrate into the SI lymph duct.
One of the most likely mechanisms responsible for this perfusion is the "blood capillary-lymph loop" ( 50 ). Our fi nding that no apo B was present in SI lymph perfusates from ABCA1 KO mice that were perfused with buffer containing serum from WT mice indicated that HDL from the systemic circulation can, whereas apo-B-containing lipoproteins cannot, fi lter into the SI lymph duct ( Fig. 1C ), suggesting that bigger molecules cannot pass through the blood-capillary wall. Consistent with this fi nding, apo B48, but not apo B100, was detected in our analyses of apos in ultracentrifugation density fractions of lipoproteins in SI lymph perfusates from WT mice ( Fig. 3B ). Therefore, we However, Duka et al. ( 56 ) showed that apo-AIV-containing HDL is formed in the absence of apo AI by using apo AI -/mice that had been transfected with the apo AIV gene. Therefore, apo AIV coexists with apo AI but can form HDL independent of apo AI if apo AI is absent. It would be interesting to know whether apo-AIV-containing HDL is formed in patients with a genetic apo AI defi ciency.
Apo AIV, which is mainly expressed in the SI, is a 46 kDa plasma protein associated with CM and HDL ( 54,57 ) and reportedly can inhibit lipid oxidation and enhance cholesterol effl ux. In addition, the overexpression of apo AIV was found to reduce atherosclerosis in mice models (58)(59)(60). Therefore, it would be of considerable interest to determine the function and relevance of SI-HDL, particularly with respect to atherosclerosis.
Because the examination of SI-HDL by EM showed spherical particles ( Fig. 5 ), we examined the involvement of ABCA1 and LCAT in the formation of SI-HDL by using inhibitors of ABCA1 and LCAT ( Fig. 6 ). We found that in SI lymph perfusates from WT mice markedly increased levels of free apo AI were detected in the presence of an ABCA1 inhibitor in the perfusion buffer, and pre-␤ -HDL appeared in the presence of an LCAT inhibitor in the perfusion buffer ( Fig. 6 ). This fi nding indicates that both ABCA1 and LCAT may be involved in the formation of SI-HDL ( 2 ).
Complete inhibition of ABCA1 or LCAT was not achieved in our experiments, and this may have been due to technical reasons; that is, the selection and dosage of inhibitors was limited because of the sensitivity of the SI to the organic solvents (methanol and ethanol) used for solving the inhibitors. However, because we have shown that SI-HDL is rich in apo AIV, our fi nding is consistent with that of Duka et al. ( 56 ), who showed that ABCA1 and LCAT participate in the biogenesis of apo-AIV-containing particles by using ABCA1 -/and LCAT -/mice that had been transfected with the apo AIV gene.
Using this novel in situ perfusion model, we also found that fasting drastically reduced, and high-fat feeding drastically increased, HDL-apo AI and HDL-C levels in SI lymph perfusates from WT mice ( Fig. 7A, B ). Our fi ndings indicate that the production of SI-HDL can be dynamically regulated by nutritional factors. It would be interesting to determine whether the ratio of apo AI in HDL to lipid-poor apo AI is similar under different dietary conditions. Our in situ perfusion model should be useful for further investigating the regulation of the production of SI-HDL by various diet components such as saturated and unsaturated fatty acids.
We found that the production of SI-HDL was markedly reduced in the major experimental murine model for atherosclerosis, apo E KO mice ( Fig. 7C ). Reduced HDL-apo AI and HDL-C in SI lymph perfusates from apo E KO mice may be caused by a redistribution of apos from HDL to non-HDL due to substantial hyperlipidemia and abnormal lipoprotein metabolism. Our results showed that apo AI in SI lymph perfusates was distributed in HDL in WT mice but was distributed in non-HDL in apo E KO mice ( Fig.  7C ). Duka et al. ( 56 ) showed that apo AIV was contained predominant in the SI lymph duct in anesthetized mice. This fi nding explains the result of Brunham et al. ( 16 ) that lymph from mice that specifi cally lack ABCA1 in the liver had no detectable HDL-C.
Our fi nding, obtained with peptide mapping of HDL using LC/MS, that C-HDL was very similar to L-HDL but different from SI-HDL ( Fig. 2 ) further confi rms our fi nding, obtained by perfusion of ABCA1 KO mice with serum from WT mice, that plasma HDL can fi ltrate from the abdominal aorta into the SI lymph duct ( Fig. 1C ). Therefore, our novel fi ndings indicate that our in situ perfusion model can selectively evaluate HDL produced from the SI without possible interference from plasma HDL or HDL derived from the liver.
Using the novel in situ perfusion model, we characterized SI-HDL in comparison with plasma HDL and L-HDL. We found that SI-HDL had a much higher protein content and a lower lipid content than plasma HDL and that CE and PL were the major lipids ( Table 1 , Fig. 4A ). Consistent with these fi ndings, by examining SI-HDL using EM, we found that most SI-HDL was spherical and HDL was smaller than plasma HDL ( Fig. 5 ). Our fi nding that SI-HDL is small and dense compared with plasma HDL suggests that SI-HDL may have higher antiatherogenic activity than plasma HDL ( 52 ).
When mice were fed ad libitum, SI-HDL separated from lymph perfusates by ultracentrifugation contained more TG and less CE than that from liver perfusates ( Table 1 ), suggesting that the composition of core lipids of intestinal HDL is different from that of hepatic HDL. It is possible that HDL becomes TG rich due to fusion between nascent HDL and TG-rich lipoprotein (TRL). We have previously shown that HDL reconstituted from apo AI and PLs remodels plasma apo-B-containing lipoprotein from a patient with Tangier disease, which was TG rich ( 53 ), and from a patient with hypercholesterolemia ( 35 ). TG in HDL is known to be hydrolyzed by hepatic TG lipase. A lack of hepatic TG lipase in SI lymph perfusates may also lead to TG-rich SI-HDL.
Our analyses of apos in ultracentrifugation density fractions of lipoproteins in SI lymph perfusates showed that apo AIV and apo AI were distributed in both the HDL and non-HDL fractions ( Fig. 3B ). We compared the compositions of apos of intestinal and hepatic HDL in HDL density fractions separated from SI lymph perfusates and liver perfusates from WT mice using ultracentrifugation ( Fig. 4B ). We found that L-HDL contained a very limited amount of apo AIV but a considerable amount of apo E, whereas an opposite trend was seen with SI-HDL ( Fig. 4B ).
Ohta et al. ( 54 ) showed that apo AIV exists as a complex with apo AI. They separated apo-AIV-containing HDL using an anti-apo AIV immunoabsorbance column from a human lymph TRL fraction, lymph lipoprotein-defi cient fraction (LDF), plasma HDL, and plasma LDF and analyzed apos after separation by SDS-PAGE. Also, Böttcher et al. ( 55 ), who separated plasma HDL into charge-based subfractions using preparative isotachophoresis, showed that slow-migrating HDL contained both apo AIV and apo AI, whereas fast-migrating HDL contained only apo AI.
Using our in situ perfusion model, we obtained novel information regarding the production of HDL by the SI, the characteristics of SI-HDL, and the regulation of HDL. However, this model is limited because the effects of anesthesia on gut motility and intestinal lipid traffi cking are not clear.
The inhibition of cholesteryl ester transfer protein (CETP) may be a strategy for raising HDL. However, it has been demonstrated that such a strategy needs to be reconsidered because some clinical trials with CETP inhibitors have failed and been terminated (61)(62)(63). We reported in the 1990s that genetic human CETP defi ciency was atherogenic rather than benefi cial (64)(65)(66). Based on both these previous and current studies, an increase in the production of SI-HDL may be a therapeutic target for raising HDL. In summary, we have shown that our in situ perfusion model using surgically isolated mouse SI achieves the selective evaluation of HDL produced from the intestine. Using this model, we showed that the production of HDL from the SI in mice requires ABCA1, and that SI-HDL is different from HDL produced by the liver and is regulated by nutritional and genetic factors. Because the intestine is a promising target for raising HDL, our in situ perfusion model represents a useful tool for developing novel strategies for the prevention and treatment of atherosclerosis. In addition, the SI performs various important functions in not only lipid homeostasis ( 67 ) but also immune defense as well as the production of hormones and cytokines. Therefore, our novel in situ perfusion system, which can be used in other spontaneous and genetically engineered mouse models, may also be a useful research tool for investigating physiological and pathological conditions in the SI and adjacent organs.