The screening assays used for this study were derived from previous work using fluorescent lipid reporters in zebrafish larvae 
. Following their ingestion, the fluorescent metabolites of these reporters are first detected in the gallbladder of live larvae and later the intestinal lumen following gallbladder contraction. The compounds are used at low concentrations and they are rapidly absorbed from the intestinal lumen, thus their fluorescence emission is not detected in the intestinal lumen immediately after ingestion or when absorption in inhibited. Fluorescence emission from one of the analogues, the phospholipid PED-6, is quenched prior to metabolism by luminal phospholipase. Thin layer chromatographic analyses of bile from adult fish, or total body lipids of 5 dpf larvae, showed that PED-6, which is labeled with a BODIPY labeled short chain fatty acid (C5) at the sn-2 position, is metabolized to cholesterol esters, phospholipids and possibly, triglyceride (19 and data not shown). Free PED6 was not detected in either assay.
For the primary screen, 5 day post-fertilization larvae were arrayed in 96 well plates and soaked overnight in test compounds (25 uM in 2% DMSO). The following morning larvae were soaked in PED-6 for 6 hours () after which a qualitative visual assessment of gallbladder fluorescence was made using an inverted compound microscope. Reduced gallbladder fluorescence, the endpoint we use to identify active compounds in the primary screen, could not differentiate compounds that inhibited lipid absorption from those that interfered with swallowing, phospholipase activity or hepatic metabolism and biliary secretion. As described below, secondary assays were devised to distinguish these mechanistic possibilities.
Chemical screen identifies novel inhibitors of zebrafish lipid metabolism.
Larvae tolerated overnight incubation in the majority of the 3,840 compounds analyzed in the primary screen, however 67 compounds (1.75%) caused larval death or severely compromised cardiac circulation and were therefore deemed toxic. 50 compounds (1.3%) caused either complete or partial inhibition gallbladder fluorescence. When re-tested in a qualitative visual assay of PED-6 metabolism, 15 of these compounds were considered active in a dose responsive fashion (0.3% of the total number screened). 12 of the 15 compounds identified in the primary screen were tested in adult fish; 5 compounds were deemed active based on reduced gallbladder fluorescence derived from PED-6 (not shown) while 7 were either inactive and or toxic in adult fish and not studied further. Together with the 3 compounds that were not available in sufficient quantity to be tested in adult fish, this left 8 compounds for testing in secondary assays (Table S1
The visual dose response assays conducted in larvae (n
5) arrayed in the 96 well plates showed that 2 of the 8 compounds first inhibited PED-6 processing at 6.25 uM (compounds 2 and 10), whereas the remaining compounds were first active at 25 uM ( shows dose response for compound 2). In separate experiments (described in the Methods
section), combined gallbladder and intestinal fluorescence of individual compound treated larvae was quantified using fluorescence microscopy. This showed that the active compounds reduced PED-6 metabolism between 51%–67% ( and Figure S1
). Of the 8 active compounds, only 1 has been used in humans; clofazimine (compound 10), a rhiminophenazine dye with antimicrobial and anti-inflammatory activity used to treat leprosy and other types of mycobacterial infections 
. Although intestinal toxicity has been reported with long term use of high doses of this drug, no prior reports of altered lipid absorption have been reported 
We devised a series of secondary assays that allowed us to further characterize the active compounds' mechanism of action and prioritize them for testing in mammals.
We assayed the effect of the active compounds on the ingestion of fluorescent microspheres to control for the possibility that they prevented swallowing of PED-6 from the larvae's aqueous media. This assay confirmed normal swallowing in 7 of 8 active compounds. Interestingly, the 1 compound that inhibited swallowing (compound 1; Figure S2
) had no obvious effect on larval motility or cardiac function.
We assayed the effect of the active compounds on the metabolism of fluorescent cholesterol and fatty acid analogues because these dietary lipids are differentially absorbed and or processed by enterocytes compared with the phospholipid used for the primary screen, PED6.
Recent studies have shown that the intestinal absorption of dietary cholesterol is dependent on the Neiman Pick Type C 1-Like 1 protein [NPC1L1; 23, 24]. Although the function of NPC1L1 is still debated, it is generally agreed upon that it as a cholesterol transporter embedded within the apical enterocyte membrane 
. NPC1L1 has not been implicated in phospholipid absorption, thus it was not predicted that the screen compounds, which were identified by their inhibition of phospholipid (PED-6) absorption, would interfere with absorption of a fluorescent cholesterol analog, NBD-cholesterol. Surprisingly, each of the 7 active compounds inhibited metabolism of NBD-cholesterol, as determined by levels of biliary and intestinal fluorescence ( and Figure S1
Cholesterol and fatty acid metabolism in zebrafish larvae treated with novel lipid absorption inhibitors.
We next measured the effect of the active compounds on the absorption of fluorescent short chain fatty acid (SCFA) and long chain fatty acid (LCFA) analogs. ( and Figure S1
). The distinction between acyl-chain length is important because LCFA are thought to be taken up from the intestinal lumen by a protein mediated process whereas as SCFA are thought to enter the enterocytes via simple diffusion 
. In addition, LCFA require incorporation into lipoprotein particles for transport from enterocytes to the liver whereas SCFA enter the blood directly and are transported bound to albumin and other serum proteins 
. All 7 compounds inhibited metabolism of the LCFA C-16 bodipy () but only 2 had an effect on SCFA C-5 bodipy metabolism (). Inhibition of native C5-bodipy processing by compounds 2 and 11 () was less pronounced than inhibition of processing of LCFA, NBD-cholesterol or PED6 (; ).
Each of the active compounds from the primary screen inhibited PED6, NBD-cholesterol and Bodipy-C16 (LCFA) metabolism. In contrast, orlistat, a pancreatic lipase inhibitor, and ezetimibe, which targets NPC1L1, are reported to inhibit absorption of only dietary 1 lipid class; triglycerides, and cholesterol and structurally related phytosterols, respectively. To determine whether the non-selectivity of the active compounds arose from a non-specific disruption of endocytic absorptive pathways in enterocytes, we assayed in vivo processing of the styryl dye AM1-43. AM1-43 is a fixable derivative of FM1-43, a reagent that has been extensively used to study endocytosis 
. When ingested by zebrafish larvae, AM1-43 strongly labels the apical plasma membrane of enterocytes. The number and size of AM1-43 labeled vesicles that can be detected in the cytoplasm of these cells provides a qualitative assessment of bulk endocytosis through the apical plasma membrane 
. 3 of the 7 active compounds (compounds 2, 7, A10) caused a marked reduction in AM1-43 processing (). Fluorescent cytoplasmic vesicles could only be detected in small percentage of the enterocytes from these larvae (n
a minimum of 10 sections from 7 compound treated and wild type larvae). The vesicles that were detected were also smaller and had lower fluorescent emission. The effect of the remaining 4 compounds (10, 11, B10, B11; and data not shown) was deemed less pronounced because a larger number of fluorescent vesicles were detected in enterocytes of treated larvae.
Endocytosis in zebrafish larvae treated with novel lipid absorption inhibitors.
To determine whether the active compounds identified in the primary screen affected other aspects of digestive physiology we assayed protease activity using a quenched bodipy labeled casein protein. Cleavage of this reporter by pancreatic proteases generates fluorescent peptides that can be detected in the intestinal lumen of wild type larvae 
. Intestinal fluorescence derived from the casein reporter was minimally reduced in larvae treated with 5 of 7 compounds (compounds 2, 10, 11, B10, B11; ). Treatment with 2 compounds (7 and compound A10) caused a profound reduction in the metabolism of the casein reporter ().
Digestive protease activity in compound treated larvae.
Changes in gallbladder and intestinal fluorescence detected in the primary screening assay detected could have arisen from a reduction in either intestinal and or hepatic lipid processing. We fed compound treated larvae egg yolk and after allowing time for its absorption, we performed whole mount stainings using the lipophilic dye oil red o (ORO) to determine whether yolk-derived lipids accumulated in either organ (). Wild type larvae fed egg yolk had strong ORO staining of the anterior intestine, as well as the blood stream, the latter arising from lipid in circulating lipoproteins 
(). Manual dissection of the intestine showed that the ORO staining derived from small lipid droplets within the enterocyte cytoplasm (data not shown). Lipid within the intestinal lumen was not detected in any wild type larvae (n
10). Each of the 7 active compounds tested reduced intestinal lipid ( and data not shown). Lipid was detected in enterocytes of all compound treated larvae, but at far lower levels than in wild type, except in larvae treated with compound 10 (). Here luminal lipid was detected. No evidence of hepatic lipid accumulation was evident. Collectively, these findings are compatible with reduced intestinal lipid absorption in compound treated larvae.
Reduced intestinal lipid absorption in ezetimibe and compound treated zebrafish larvae.
Compound synergy was examined in binary combinations of the 7 remaining active compounds with each other and with ezetimibe. Each compound was assayed at the highest concentration deemed inactive and the lowest dose considered active in the visual dose response experiment. These experiments identified potential synergism between compounds 2 and 10 (data not shown).
The two most commonly prescribed lipid absorption inhibitors, orlistat and ezetimibe, are generally considered to be selective inhibitors of triglyceride, and cholesterol and phytosterol absorption, respectively. To gain a better understanding of the mechanism of action of the novel compounds identified in our screen, we examined how these drugs affected absorption of fluorescent lipid reporters in zebrafish larvae.
Both drugs were assayed in an identical fashion as the screen compounds. Orlistat had no effect on the metabolism of any of the lipid reporters (data not shown). This was predicted, however because none are processed by pancreatic lipase, which is responsible for hydrolysis of triglycerides 
. In contrast to orlistat, ezetimibe was predicted to inhibit absorption of NBD-cholesterol because the amino acid domain of dog Npc1l1 required for high affinity binding to ezetimibe 
is highly conserved in both human NPC1L1 and the predicted zebrafish Npc1l1 protein (42 identical, 11 conserved and 10 non-conserved residues; Figure S3
). Indeed, larvae treated overnight with the highest ezetimibe dose tested (50 uM) showed a 78% reduction of gallbladder and intestinal fluorescence derived from NBD-cholesterol (). Treatment with lower doses (37.5 uM, 25 uM, 12.5 uM, 6.25 uM) showed proportionately less inhibition (Figure S4
). Unexpectedly, ezetimibe also reduced metabolism of the phospholipid PED-6 and the saturated long chain fatty acid (LCFA) Bodipy-C16 (). As predicted, ezetimibe had minimal effect on the metabolism of SCFA Bodipy-C5 (). ORO stainings of yolk-fed larvae confirmed reduced lipid absorption was reduced by ezetimibe treatment (). Ezetimibe had no effect on digestive protease function in zebrafish larvae (data not shown).
Ezetimibe inhibits lipid metabolism in zebrafish larvae.
Previous work suggests that ezetimibe interferes with intestinal cholesterol absorption by disrupting the incorporation of NPC1L1 into clathrin-coated vesicles 
. This mechanism does not predict that ezetimibe will interfere with fatty acid or phospholipids uptake by enterocytes, neither of which are known to be dependent on NPC1L1. Because of this, we speculated that ezetimibe had a broader disruptive effect on intestinal endocytic mechanisms. To examine this, we measured uptake of AM1-43 in ezetimibe treated larvae. Compared with control larvae, ezetimibe (50 uM) treated larvae had a markedly reduced number of AM1-43 labeled vesicles in enterocytes of the anterior intestine, the site of lipid absorption in zebrafish larvae (). The effect of ezetimibe on AM1-43 uptake was dose responsive ().
Membrane cholesterol depletion and ezetimibe inhibit endocytosis and fatty acid metabolism in zebrafish larvae.
To gain additional insight into the mechanism of action of ezetimibe as well as the active compounds that affected endocytosis (2, 7, A10), we compared their effect on AM1-43 metabolism with the effect of methyl-β-cyclodextrn (MβC), a reagent that disrupts membrane lipid rafts by extracting membrane cholesterol 
. Pretreatment of zebrafish larvae with MβC for four hours strongly inhibited endocytic uptake of AM1-43 by enterocytes (). Recovery of endocytic function was detected eight hours after MβC withdrawal (), but was prevented in larvae unable to replenish membrane cholesterol because of concomitant treatment with the cholesterol synthesis inhibitor atorvastatin (). Atorvastatin treatment on its own had no effect on AM1-43 processing (). Like ezetimibe and the compounds that interfered with AM1-43 processing, MβC inhibited C-16 bodipy metabolism, and this too was reversed by repletion of membrane cholesterol (). MβC had minimal effect on C-5 bodipy metabolism (), most likely because enterocytes absorb SCFA via passive diffusion.