Trans-intestinal cholesterol efﬂ ux is not mediated through high density lipoprotein

absorbed the gut into chylomicrons secretes Abstract Transintestinal cholesterol efflux (TICE) pro-vides an attractive target to increase body cholesterol excretion. At present, the cholesterol donor responsible for direct delivery of plasma cholesterol to the intestine is unknown. In this study, we investigated the role of HDL in TICE. ATP-binding cassette protein A1 deﬁ cient ( Abca1 (cid:2) / (cid:2) ) mice that lack HDL and wild-type (WT) mice were intravenously injected with chylomicron-like emulsion particles that contained radiolabeled cholesterol that is liberated in the liver and partly reenters the circulation. Both groups secreted radiolabeled cholesterol from plasma into intestinal lumen and TICE was unaltered between the two mouse models. To further investigate the role of HDL, we injected HDL with radiolabeled cholesterol in WT mice and Abca1 (cid:2) / (cid:2) ×Sr-b1 (cid:2) / (cid:2) mice that lack HDL and are also unable to clear HDL via the liver. The intestines of both mice were unable to take up and secrete radiolabeled cholesterol from HDL via TICE. Although a generally accepted major player in the hepatobiliary route-based cholesterol excretion, HDL plays no signiﬁ cant role in TICE in mice. —Vrins, C. L. J., R. Oever, Trans-intestinal cholesterol efﬂ ux is not mediated through high density lipoprotein. • apolipoproteins • reverse cho- lesterol transport • neutral sterols • bile

37°C in a shaking-water bath under argon. Ethylmercurithiosalicylate (thimerosal; 20 mM) was added to inhibit phospholipid transfer and lecithin:cholesterol acyltransferase activity. Radiolabeled HDL was then isolated by density gradient ultracentrifugation.

Intestine perfusions, liver uptake, and plasma decay
Mice (2-4 months old, n = 5) were anesthetized by intraperitoneal injection of a mixture of 7 ml fl uanisone (17.5 mg), fentanyl citrate (0.55 mg), and midazolam (8.75 mg) per kg body weight. Continous anesthesia was monitored by tail pinching. Radiolabeled chylomicron-like particles ( H]CO/mouse) were administered by tail vein injection. Plasma samples were collected at the indicated time points after injection through tail-bleeding. After 1 h, bile was collected via bile cannulation for 15 min and the proximal small intestines (fi rst 10 cm) were perfused with a bile salt-phospholipid mixture that was composed of 10 mM taurocholate and 2 mM phosphatidylcholine (PC) and at a fi xed fl ow rate of 3 ml/hr, as described previously ( 5 ). This mixture of 10 mM taurocholate and 2 mM PC was shown earlier to yield optimal rates of cholesterol secretion ( 5 ). At the end of the perfusion period of 90 min, blood was collected by cardiac puncture. Liver was snap-frozen in liquid N 2. The total amount of radioactivity in the plasma was calculated based on the estimated total plasma volume (4.5% of body weight) ( 16,17 ). To determine liver uptake, the livers were weighed and tissue samples were dissolved in Soluene-350 (PerkinElmer, Waltham, MA). To determine intestinal uptake, the perfused intestinal segments were harvested; the mucosa was scraped, snap-frozen in liquid N 2 , and lipids were extracted according to Bligh and Dyer ( 18 ). For all samples radioactivity was determined using a liquid scintillation counter.

TLC
To determine the distribution of 3 H and/or 14 C over free cholesterol and cholesteryl esters in homogenized samples of snapfrozen livers and plasma samples, lipids were extracted according to Bligh and Dyer ( 18 ). After reconstitution of lipid fi lms in chloroform, free cholesterol and cholesteryl esters were separated by spotting the samples on silica gel 60 TLC plates (Merck, Darmstadt, Germany). Subsequently, the sterols were identifi ed by iodine staining and the distribution of radioactivity was quantitated using a liquid scintillation counter.

Cholesterol measurements
Biliary and perfusate cholesterol were measured by a fl uorescent method as described previously ( 19 ). Total cholesterol in plasma was determined using cholesterol RTU kit from Biomerieux (Marcy l'Etoile, France).

Statistical analysis
All results are presented as mean ± SD. Group means for TICE, as depicted in the fi gures, were calculated by averaging the outcomes of all mice that got the same treatment. The values for the individual mice, used for the calculation of the group mean, were obtained by averaging the data of the last three perfusion time points. Differences between different groups were determined by one-way ANOVA. Outcomes of P < 0.05 were considered to be signifi cant.

Cholesterol introduced via chylomicron-like emulsion particles is secreted via TICE
In our attempt to identify the cholesterol donor for TICE, we fi rst investigated the fate of [ 3 H]CO and [ 14 C] these lipoprotein particles into lymph for distribution over the body ( 7 ). In addition, the intestine is an important contributor to circulating HDL levels ( 8 ). However, little is known about uptake of circulating lipoproteins by the intestine from the basolateral side. Interestingly, the rate of TICE was two-fold higher in SR-BI-defi cient mice with increased HDL levels as related to an impaired delivery of HDLcholesterol to the liver ( 9 ). It is thus conceivable that HDL is utilized for RCT via the hepatobiliary route as well as TICE.
In this study, we investigated basolateral uptake of cholesterol by the intestine and the potential role for HDL in TICE. We evaluated plasma kinetics, liver uptake, and biliary and intestinal secretion of radiolabeled cholesterol, which was injected via chylomicron-like particles in mice with normal and disturbed HDL metabolism. Results reported herein indicate that TICE is not mediated significantly via HDL particles.

Preparation of chylomicron-like emulsion particles
Chylomicron-like emulsion particles were prepared as described ( 13 ). Briefl y, a mixture of 100 mg of total lipid was dispersed in NaCl buffer of density 1.10 g/ml. The lipid mixture consisted of egg yolk phosphatidylcholine (Lipoid, Ludwigshafen, Germany), triolein, L-␣ -lysophosphatidylcholine, cholesteryl oleate, and cholesterol (all from Sigma-Aldrich, The Netherlands) and at a weight ratio of 22

Preparation radiolabeled HDL
Human HDL was isolated from blood of healthy subjects by differential ultracentrifugation as described by Redgrave et al. ( 14 ) and dialyzed against PBS with 1 mM EDTA. HDL (1.063 < d <1.21 g/ml) was labeled with [ 3 H]CO via exchange from donor particles as reported previously ( 15 ). Donor particles were formed by sonication of egg yolk phosphatidylcholine supplemented with 50 µCi of [ To determine TICE and biliary cholesterol secretion, we performed intestine perfusions and bile cannulations, starting at 1 h after injection, for a period of 90 min. Table 1 shows that secretion of cholesterol via TICE was 2.5-fold higher than the biliary cholesterol secretion. This fi nding confi rmed our earlier observation that TICE plays a very prominent role in fecal neutral sterol excretion in mice ( 5,20 ). A substantial amount of the intravenously injected radiolabeled cholesterol taken up by the liver is cleared via biliary cholesterol secretion. In addition, we also demonstrate with our intestine perfusions that radiolabeled cholesterol is secreted into the intestinal lumen ( Table 1 ). In all measured compartments, the activity ratio between 3 H and 14 C remained approximately 4:1 (data not shown). Although these data confi rm direct secretion of cholesterol from blood into the intestinal lumen via TICE, they do not discriminate between different potential cholesterol donors and previous work already excluded direct uptake of our radiolabeled emulsion particles by the intestine ( 20 ).

Secretion of cholesterol from chylomicron-like emulsion particles via TICE is relatively unaffected in mice that lack HDL or HDL uptake
In a previous report, we showed that Abca1 Ϫ / Ϫ mice with almost no HDL levels show unaltered fecal neutral sterol excretion compared with their WT littermates ( 24 ). To investigate whether TICE is also unaltered in the absence of HDL, we investigated the transport of radiolabeled cholesterol from chylomicron-like emulsion particles into the intestinal lumen. To this end, we injected [ 3 H] CO-labeled particles into Abca1 Ϫ / Ϫ mice and their WT littermates. Clearance of 3 H-activity from plasma and hepatic uptake of 3 H-activity in Abca1 Ϫ / Ϫ mice was similar to that in cholesterol contained in chylomicron-like emulsion particles upon administration to WT mice by intravenous injection. Both [ 3 H]CO and [ 14 C]cholesterol were cleared rapidly from the circulation (t 1/2 = 6-7 min) ( Fig. 1A ). In previous reports we have shown that, in rats and mice, the majority of the injected dose (>60%) is taken up by the liver within 30 min and that no direct uptake is observed by the intestine ( 13 ). After 2.5 h, 30% of the injected dose was still present in the liver, which was similar to our previous observation and earlier reports ( 14,20 ). Interestingly, this was observed for both isotopes (Fig. 1B), suggesting that no exchange occurs of free cholesterol (and CO) from the emulsion particles to their environment in plasma before uptake by the liver.
To analyze the distribution of both the 3 H and the 14 C radiolabel between its free and esterifi ed form in the liver, TLC was performed on extracted lipids of liver homogenates (Fig. 1C). Although the introduced [ 14 C]cholesterol remained largely in the free form, the majority of introduced [ 3 H]cholesteryl oleate was hydrolyzed and also present as free cholesterol. In plasma of these mice, cholesteryl esters made up about 75% of the total circulating cholesterol. Analysis of plasma 2.5 h after injection showed that the 3 H activity in the circulation (9.1 ± 1.5% of the injected dose) was evenly distributed over free and esterifi ed cholesterol (Fig. 1D). A similar distribution was found for plasma 14 C activity (9.7 ± 1.6% of injected dose). These data illustrate the reutilization of the radiolabeled cholesterol after specifi c uptake by the liver as demonstrated before (21)(22)(23). In addition, lipoprotein analysis of these plasma samples by fast protein liquid chromatography demonstrated that, like the total plasma cholesterol, both radiolabels were mostly associated with the HDL fraction (Fig. 1E).   Fig. 2A (right panel), the label was cleared rapidly from the circulation and there was no difference between WT and knockout mice. Interestingly, specifi c activity of cholesterol was unaltered both in bile and intestinal perfusate derived from Sr-b1 Ϫ / Ϫ mice (Fig. 3). Ϫ / Ϫ mice ( Fig. 4A ).

Cholesterol introduced via HDL is not secreted via TICE
In addition, the lowered hepatic uptake of [ 3 H]CO is refl ected by the reduced specifi c activity of cholesterol WT littermates ( Fig. 2A , B; left panels). Interestingly, despite the absence of HDL, specifi c activity of biliary cholesterol secretion was not signifi cantly altered ( Fig. 3A ; left panel), in line with the unaltered biliary cholesterol secretion previously reported by our group ( 24 ). Intestine perfusions performed on the Abca1 Ϫ / Ϫ mice showed that TICE was not signifi cantly changed. As monitored by the appearance of unlabeled cholesterol in the perfusate samples, the rate of TICE in mice lacking HDL was 1.4 ± 0.5 nmol/min.100g body weight compared with 1.9 ± 0.6 nmol/min.100g body weight in their WT littermates. Furthermore, despite the absence of HDL in Abca1 Ϫ / Ϫ mice, transintestinal secretion of radiolabeled cholesterol introduced via chylomicron-like emulsion particles was still 60% compared with WT littermates (Fig. 3B; left panel). Mice with a disrupted Sr-b1 gene show elevated HDL and have increased TICE ( 9 ). To investigate whether   Because HDL plays an important role in RCT via the hepatobiliary route and contains the majority of circulating lipoprotein cholesterol in mice, we considered HDL a likely candidate to directly supply the intestine with cholesterol for TICE. Consistent with this hypothesis, in Sr-b1 Ϫ / Ϫ mice that are characterized by elevated HDLcholesterol levels, we found a two-fold increase in TICE ( 9 ); however, as demonstrated in the present study (Fig.  3B), specifi c activity in the perfusate was unaltered, which may be due to the increased pool of free cholesterol in the plasma ( 11 ). We now show also that in Abca1 Ϫ / Ϫ mice without detectable HDL-cholesterol, TICE is not altered compared with their WT littermates. In fact, we still detected a signifi cant secretion of radiolabeled cholesterol from injected chylomicron-like emulsion particles via TICE. The specifi c activity of the cholesterol secreted via TICE in Abca1 Ϫ / Ϫ mice was approximately 60% of that found in their WT littermates. It cannot be excluded that the lack of TICE phenotype in Abca1 Ϫ / Ϫ mice is due to compensatory mechanisms inducing peripheral cholesterol effl ux in an HDL independent manner. In addition, possible decreased pheripheral cholesterol fl ux could be compensated for by increased cholesterol de novo synthesis in the intestine. HDL as a critical cholesterol donor for TICE was further refuted after we introduced [ 3 H]CO-labeled HDL into mice. As expected, clearance of HDL via the liver was impaired in Abca1 Ϫ / Ϫ ×Sr-b1 Ϫ / Ϫ mice due to Sr-b1 defi ciency ( 11 ). Importantly, no uptake of HDL by the intestine was observed and hardly any radiolabeled cholesterol could be detected in the perfused intestinal lumen. Interestingly, no uptake and secretion of HDLderived cholesterol could be detected in the WT littermates. These data suggest not only that TICE is not mediated via HDL, but also that little or no uptake of HDL by the intestine takes place, confi rming earlier studies by Briand et al. ( 29 ). If HDL at a physiologic concentration is secreted into bile (Fig. 4B). However, biliary cholesterol secretion was unaltered in the Abca1 Ϫ / Ϫ × Sr-b1 Ϫ / Ϫ mice ( 12 ). Whereas Sr-b1 Ϫ / Ϫ mice were characterized by an elevated cholesterol secretion by the intestine ( 9 ), the rate of TICE (in nmol/min.100g body weight) in the double knockout mice was similar compared with their WT littermates (Fig. 4C). Interestingly, HDL-derived cholesterol was hardly secreted by the intestine in both Abca1 Ϫ / Ϫ × Sr-b1 Ϫ / Ϫ and WT mice, indicating that HDL is not taken up by the intestine and, concomitantly, hardly any radiolabeled cholesterol incorporated in the injected HDL could be secreted via TICE (Fig. 4D).

DISCUSSION
The existence of TICE has been demonstrated in various studies in mouse models ( 5,25 ) and earlier results indicate that this pathway is also present humans ( 26 ). The contribution of direct cholesterol secretion via the intestine to fecal sterol excretion is a novel concept that can advance the development of cholesterol-lowering therapies. However, the mechanism of TICE remains to be elucidated. An important question that needs to be answered is which donor particle delivers the cholesterol for secretion via TICE. In the RCT pathway via the hepatobiliary route, HDL plays an important role ( 27 ). In this report, we provide important evidence that HDL does not contribute signifi cantly to TICE.
Our results from intestine perfusions show that upon injection of chylomicron-like emulsion particles, as a model of endogenously circulating triglyceride-rich lipoproteins ( 17 ), the incorporated radiolabeled cholesterol can be secreted directly into the intestinal lumen independent of the biliary route. It was shown previously that these chylomicron-like emulsion particles acquire lipoproteins in the circulation and are rapidly taken up rapid by the liver ( 13,20 ) via apoE-selective receptors. Confi rming the landmark study of Robins and Fasulo ( 28 ), that part of the in the liver-liberated cholesterol is preferentially secreted into bile and part of is resecreted into the circulation and is distributed over the lipoproteins including HDL, similar  preferentially takes the hepatobiliary route but when abrogated, the cholesterol can be resecreted from the liver and be excreted via TICE.

Limitations of this study
The integrative physiological approach used in this study does not allow a detailed probe into the mechanism by which cholesterol is donated to the enterocyte. We have attempted to set up a transcellular cholesterol transport study in cultured CaCo2 cells but failed to visualize a TICElike transport (data not shown). The results reported here cannot be directly transplanted to the human situation. Mice lack CETP and have a much lower biliary cholesterol relative to bile salt secretion compared with humans (33 ). Lack of CETP can be overcome by carrying out future studies in CETP transgenic mice. Humanization of biliary lipid secretion is more diffi cult to realize.
Despite these limitations, we feel this data allows for the conclusion that direct secretion of cholesterol from blood through the intestine into the intestinal lumen of mice occurs independently of HDL and bile. However, the identity of the cholesterol donor responsible for delivery of cholesterol to the intestine for secretion via TICE remains to be elucidated. With the intestine being an accessible target, the identifi cation of this step in the mechanism of TICE would benefi t the development of novel and less invasive cholesterol-lowering therapies.
apparently not involved in TICE, it is interesting to speculate why we observed a two-fold increase of TICE in Sr-b1 Ϫ / Ϫ mice in an earlier study ( 9 ). A closer look at the lipoprotein profi le of these mice shows not only an increase of the HDL-peak but also a shift of HDL toward larger lipoprotein particles and an overlap with the LDL/IDL fractions ( 30 ). Furthermore, the abnormally large HDL particles that accumulate in SR-BI-knockout mice are enriched in apoE ( 11 ). It is thus possible that the severe enrichment of HDL with apoE caused the increase in TICE and may indicate that apoE may mediate TICE from lipoprotein remnants. Due to the extremely rapid metabolism of VLDL and LDL, we have not been able to delineate whether these particles specifi cally serve as substrate for TICE.
In an attempt to distinguish between the fate of free and esterifi ed cholesterol, we incorporated both [ 14  ]cholesterol, it remained in the free form. Part of this hepatic radiolabeled cholesterol pool is resecreted into the circulation and distributed among the lipoproteins that circulate in WT mice. During the entire timeline from injection to secretion into the intestinal lumen, the ratio between the two isotopes ( 14 C: 3 H = 1: 4) remained similar. However, because both isotopes upon redistribution by the liver were then evenly distributed over free and esterifi ed cholesterol, it was not possible to determine a possible preference by the intestine for either of the cholesterol forms.
Based on observational and epidemiological studies, increased HDL levels are considered to be clinically beneficial partly due to an improved RCT via the hepatobiliary route ( 31 ). However, our study indicates that HDL apparently does not play a role in TICE. When, as generally accepted, HDL is the primary acceptor for uptake of excess cholesterol from macrophages, TICE may play no role in the pathogenesis of atherosclerosis. A recent study by Temel et al. ( 6 ) disproves the selectivity of HDL in macrophagespecifi c RCT. When acutely bile diverted mice were intraperitoneally injected with cholesterol-loaded macrophages, the radiolabled cholesterol appeared undeterred in the intestinal lumen demonstrating that TICE can mediate macrophage specifi c RCT. In contrast, Nijstad et al. ( 32 ) reported that mdr2 knockout mice with abrogated biliary cholesterol secretion showed greatly reduced fecal excretion of macrophage derived radioactive cholesterol. Very recently, Zhao et al. ( 12 ) used the same methodology to determine RCT in the animal models used in the present study. Here, it was observed that RCT from WT macrophages was inhibited about 25% in both Abca1 Ϫ / Ϫ mice and Abca1 Ϫ / Ϫ × Sr-b1 Ϫ / Ϫ mice. Hence, upon full disruption of hepatic HDL-mediated cholesterol uptake the majority of macrophage-specifi c RCT still continues. These data support the conclusions of Temel