|Home | About | Journals | Submit | Contact Us | Français|
Hepatic stellate cells (HSCs) are liver-specific mesenchymal cells that play vital roles in liver development and injury. Our knowledge of HSC biology is limited by the paucity of in vivo data. HSCs and sinusoidal endothelial cells (SECs) reside in close proximity and interactions between these two cell types are potentially critical for their development and function. Here we introduce a transgenic zebrafish line, Tg(hand2:EGFP), that labels HSCs. We find that zebrafish HSCs share many similarities with their mammalian counterparts, including morphology, location, lipid storage, gene expression profile, and increased proliferation and matrix production in response to an acute hepatic insult. Using the Tg(hand2:EGFP) line, we conducted time course analyses during development to reveal that HSCs invade the liver after SECs do. However, HSCs still enter the liver in mutants that lack most endothelial cells including SECs, indicating that SECs are not required for HSC differentiation or their entry into the liver. In the absence of SECs, HSCs become abnormally associated with hepatic biliary cells, suggesting that SECs influence HSC localization during liver development. We analyzed factors that regulate HSC development and show that inhibition of vascular endothelial growth factor signaling significantly reduces the number of HSCs that enter the liver. We also performed a pilot chemical screen and identified two compounds that affect HSC numbers during development.
Our work provides the first comprehensive description of HSC development in zebrafish and reveals the requirement of SECs in HSC localization. The Tg(hand2:EGFP) line represents a unique tool for in vivo analysis and molecular dissection of HSC behavior.
Hepatic stellate cells (HSCs) represent a versatile mesenchymal cell type that plays vital roles in liver function and injury response. In healthy livers, HSCs serve as the main vitamin A-storing cells. Upon liver injury, these quiescent cells transform into activated, proliferative myofibroblast-like cells to generate scar tissue (reviewed by 1). Sustained activation of HSCs is a central event in liver fibrosis and has been linked to the progression of hepatitis and steatohepatitis (reviewed by 2).
Despite the importance of HSCs in liver physiology and disease, our knowledge of HSC biology is far from complete. Characterization of HSC development could provide clues for understanding their activation during liver injury. However, tracking HSCs during development has not yet been feasible in mammalian models. The embryonic origin of HSCs is elusive because they express marker genes of all three germ layers (3, 4). Genetic lineage-tracing analysis in mice using a CreERT2 transgene knocked into the Wilms’ tumor suppressor (Wt1) locus indicated that HSCs derive from the septum transversum-derived mesothelium (5). Interestingly, in chick, the mesothelium contributes not only to HSCs, but also to sinusoidal endothelial cells (SECs) (6). HSCs and SECs exhibit close physical association and common expression of angiogenic factors (7). These observations have led to the hypothesis that HSCs and SECs originate from a common embryonic precursor. However, no studies have specifically addressed this issue.
Finding promoters that selectively drive transgene expression in HSCs could facilitate both their in vivo observation and their genetic manipulation. Previous studies used the promoters of the mesoderm-associated α-smooth muscle actin gene and the neural crest-related glial fibrillary acidic protein/GFAP gene to direct gene expression in HSCs in transgenic mice (8-10). However, identifying HSC-specific promoters remains challenging. The bHLH transcription factor gene heart and neural crest derivatives expressed transcript 2/hand2 is expressed in the lateral plate mesoderm and the neural crest (11), tissues that may give rise to HSCs. We previously reported the generation of the Tg(hand2:EGFP) zebrafish line that expresses EGFP under the control of the hand2 regulatory sequences (12). During liver budding morphogenesis, Tg(hand2:EGFP) is expressed in the lateral plate mesoderm surrounding the liver primordium. However, Tg(hand2:EGFP) expression was not characterized during later stages of liver development.
The teleost zebrafish (Danio rerio) has emerged as a valuable vertebrate model system for studying liver development and disease. The zebrafish liver contains the same main cell types as the mammalian liver, including hepatocytes, biliary cells, and endothelial cells (13). Although the basic architecture of the fish liver differs from that of the mammalian liver, mechanisms of liver development and diseases are conserved (reviewed by 14). In addition, the zebrafish model presents important advantages that complement those of other animal models. The rapid external development and translucence of the embryos and larvae make them well suited for in vivo imaging analyses. Using transgenic approaches, investigators have generated zebrafish that express fluorescent proteins in different hepatic cell types, allowing easy visualization of hepatic cell behaviors in the animal, and greatly facilitating genetic and chemical screens to identify regulators of liver development and disease pathogenesis (reviewed by 14). Although intensive studies have been conducted on parenchymal cells in the zebrafish liver, no report has yet focused on HSCs.
In this study, we report that the Tg(hand2:EGFP) line marks HSCs during both embryonic and adult stages. Zebrafish HSCs share significant similarities with mammalian HSCs, including their morphology, localization, lipid storage, and gene expression. They respond to acute alcohol exposure by changing morphology, upregulating extracellular matrix protein production and increasing proliferation. By tracking HSCs throughout development, we show that zebrafish HSCs enter the liver after SECs do. Study of cloche/clo mutants which lack SECs indicates that although SECs are not required for HSC differentiation or their entry into the liver, they influence the localization of HSCs inside the liver. We also reveal that inhibition of vascular endothelial growth factor (VEGF) signaling impairs entry of HSCs into the developing liver. Taken together, our work presents a new in vivo model to study HSC biology and provides novel insights into the molecular and cellular mechanisms underlying HSC development.
Wild-type, clos5+/-, Tg(hand2:EGFP)pd24, Tg(kdrl:ras-mCherry)s896, Tg(fabp10a:dsRed)gz15 strains were maintained as described (15). The genotype of clos5-/- embryos was determined by the lack of blood cells and severe edema (16). The University of California at San Francisco Institutional Animal Care and Use Committee approved all protocols.
Methods for these experiments are described in the Supporting Information.
Acute ethanol treatment was conducted as described (17). To monitor the behaviors of HSCs after ethanol treatment, larvae were transferred back to embryo medium immediately after treatment and put on regular hatch fry diet.
To assess HSC proliferation during ethanol treatment, Tg(hand2:EGFP) larvae were incubated in 7 μM 5-ethynyl-2’deoxyuridine (EdU) dissolved in embryo medium with or without 2% ethanol for 24 hours. To assess HSC proliferation after treatment, control and ethanol-treated larvae were removed from ethanol and incubated in EdU solution for 24 hours. Animals were processed using the Click-iT EdU Imaging Kit (Invitrogen).
Tg(hand2:EGFP;kdrl:ras-mCherry) animals were treated with 1 or 2 μM SU5416 (Sigma) in embryo medium at the stages indicated. Control animals from the same batch were treated with equal concentrations of DMSO. Microinjection of kdrl morpholino was performed as described (18). To quantify the number of intrahepatic vascular branches, three-dimensional projections were obtained from confocal stacks scanning through the entire liver. Each vascular branch was outlined using the Paintbrush tool in ImageJ and the number of branches was counted. For HSCs, the number of Tg(hand2:EGFP)-expressing cells located inside the liver was counted. Those cells that were still closely associated with the liver periphery were excluded. Statistical analyses were performed using the Student's two-tailed t test.
Materials and methods for these experiments are described in the Supporting Information.
Consistent with the expression of endogenous hand2 mRNA (Fig. 1A) (11), Tg(hand2:EGFP) is expressed in the neural crest and the lateral plate mesoderm and their derivatives during development (Fig. 1A’). Interestingly, we detected sparse expression of Tg(hand2:EGFP) within the liver. To determine the identity of these Tg(hand2:EGFP)-expressing cells, we examined their morphology and distribution using confocal microscopy. By 4 days post fertilization (dpf), Tg(hand2:EGFP) is expressed in two distinct cell populations associated with the liver (Fig. 1B): one population forms a single cell layer lining the liver surface (arrows), whereas the other is located within the liver (asterisks). The cells inside the liver do not express molecular markers of hepatocytes (Tg(fabp10a:dsRed)) (19) (Fig. 1C), biliary cells (Alcam) (13) (Fig. 1D), or endothelial cells (Tg(kdrl:ras-mCherry)) (20) (Fig. 1E). They display a star-like configuration and form complex cellular processes that appear to wrap around endothelial cells (Fig 1E, arrows), features that are characteristic of HSCs (1). We also detected Tg(hand2:EGFP)-expressing cells in the adult liver where they exhibit similar morphology and close association with SECs (Fig. 1F).
To determine whether the Tg(hand2:EGFP)-expressing cells are indeed HSCs, we stained the animals with antibodies that recognize the HSC markers GFAP (3) (Fig. 2A, A') and desmin (4) (Fig. 2B, B'). At 128 hours post fertilization (hpf), we detected on average 55 Tg(hand2:EGFP)-expressing cells in the liver (8 embryos analyzed). 90% of these cells were labeled by the desmin antibody and 84% of them by the GFAP antibody. Thus Tg(hand2:EGFP) expression largely overlaps with HSC marker labeling. In mammals, HSCs serve as the main vitamin A-storing cells in the body. We performed gold chloride staining that labels retinoids (21) and found that the adult zebrafish liver stores vitamin A droplets (Fig. 2C). We also detected lipid droplets inside the Tg(hand2:EGFP)-expressing cells by Oil Red O staining (22) (Fig. 2D, asterisks). These data strongly suggest that Tg(hand2:EGFP) expression marks HSCs in zebrafish.
We isolated Tg(hand2:EGFP)-expressing cells from adult zebrafish livers via fluorescence-activated cell sorting (FACS), and performed gene-profiling analysis to detect transcripts that exhibit high expression levels in HSCs, but baseline expression levels in other hepatic cells (Supporting Fig. S1A). Among the most differentially expressed transcripts, we identified genes that have been previously implicated in mammalian HSC biology (Table S1; Supporting Fig. S1B-C). Hence, zebrafish HSCs exhibit a similar gene expression profile as their mammalian counterparts.
To determine the functional relevance between zebrafish and mammalian HSCs, we assessed the response of zebrafish cells to a hepatic insult. We selected ethanol as a stimulus, because alcoholic liver disease is an important cause of HSC activation and liver fibrosis (23), and because zebrafish represent an excellent model for studying the effects of alcohol on the liver (17). We exposed Tg(hand2:EGFP) larvae to 2% ethanol from 96 to 120 hpf and monitored the behaviors of HSCs during and after treatment. All the animals survived acute ethanol exposure but showed body abnormalities and erratic swimming behaviors (300 larvae from six clutches were examined). 60% of the treated animals developed steatosis, consistent with a previous report (17). We examined the deposition of matrix proteins in untreated control and ethanol-treated livers. Whereas laminin was almost undetectable in control livers (Fig. 3A’), its deposition was markedly elevated in ethanol-treated livers (Fig. 3B’), suggesting that acute ethanol exposure stimulates matrix deposition by zebrafish HSCs. Similar to laminin, type IV collagen was also deposited in excess in ethanol-treated livers (data not shown). Furthermore, the morphology of HSCs was altered after ethanol treatment: they lost the complex cytoplasmic processes and their cell bodies became more elongated (Fig. 3B).
When ethanol-treated animals were transferred back to embryo medium, their body phenotypes and abnormal swimming behaviors recovered within a day and more than 70% of the animals survived for at least one week. During the followup period after ethanol removal, HSCs rapidly increased in number in ethanol-treated animals compared to controls (Fig. 3C-E). To determine whether cell proliferation contributed to this increase in HSC number, we examined incorporation of the proliferation marker EdU by these cells. In untreated animals, 20% of the HSCs showed EdU incorporation and this percentage was unchanged during ethanol treatment (Fig. 3F). However, one day after treatment, approximately 40% of the HSCs in ethanol-treated larvae incorporated EdU, compared to 28% in controls (Fig. 3F). Thus HSCs became more proliferative after acute ethanol treatment, which was at least partially responsible for the increase in HSC cell number in treated livers.
Taken together, our observations indicate that zebrafish HSCs exhibit enhanced matrix protein deposition, morphological changes, and increased proliferation upon acute alcohol exposure. They support the concept that zebrafish HSCs are functionally similar to their mammalian counterparts during liver injury.
The process of HSC development is incompletely understood. The close association between HSCs and SECs suggests that SECs may play a role in HSC development. We performed time course analyses to monitor the interactions between HSCs and SECs during development. To visualize SECs, we used the Tg(kdrl:ras-mCherry) line, in which the promoter of the VEGF-receptor gene kdrl drives ras-mCherry expression (20). We detected strong Tg(kdrl:ras-mCherry) expression in SECs but not in HSCs, enabling us to distinguish between these two cell populations in Tg(hand2:EGFP;kdrl:ras-mCherry) animals.
Early zebrafish liver development proceeds without SECs, which do not enter the liver prior to 55 hpf (13). Between 62 and 64 hpf, SECs are situated mostly at the dorsal surface of the liver (Fig. 4A, B, arrows), while Tg(hand2:EGFP)-expressing cells are restricted to the boundary between the liver and the gut (Fig. 4A, B, asterisks). By 66 hpf, SECs begin to invade the liver (Fig. 4C, arrow) (13), while Tg(hand2:EGFP)-expressing cells remain at the liver periphery (Fig. 4C, asterisks). From 68 hpf onwards, HSCs gradually spread throughout the liver (Fig. 4D-F). Both HSCs and SECs seem to enter the liver at random locations. Whereas some SECs invade the liver without being accompanied by HSCs (Fig. 4D, arrow), all HSCs maintain close proximity to SECs once they are inside the liver (Fig. 4D, arrowheads).
While tracking HSC development, we noticed that the number of HSCs increased from 5±4 to 33±10 (average±standard deviation) between 65 and 81 hpf. To determine whether proliferation of HSCs accounted for the increase in HSC number, we performed immunohistochemistry with the anti-Phospho-histone 3 antibody to label M-phase cells (24) (Fig. 4G-J). In all the animals examined, we detected 100 Phospho-histone 3 and Tg(hand2:EGFP) double positive cells in the liver. However, only eight of these cells resided inside the liver (Fig. 4G, arrow), while the remainder were located at the liver periphery (Fig. 4H-J, arrows). These results suggest that Tg(hand2:EGFP)-expressing cells proliferate mainly at the liver periphery prior to entering the liver.
To investigate the role of SECs in HSC development, we examined Tg(hand2:EGFP) expression in embryos homozygous for the clo mutation, which lack most hematopoietic and endothelial cells (16). At 4 dpf, whereas wild-type livers formed a rudimentary intrahepatic vascular network as revealed by Tg(kdrl:ras-mCherry) expression (Fig. 5A, A’), we did not detect any Tg(kdrl:ras-mCherry) expression in clo mutant livers (Fig. 5B, B’). It has been shown that in zebrafish, endothelial cells are not required for liver budding or hepatocyte differentiation (25), but they appear to be essential for further growth of the liver (26). Consistent with these reports, clo mutant livers contained only two-thirds as many Prox1+ parenchymal cells as wild-type livers at 4 dpf (Fig. 5E). However, the number of Tg(hand2:EGFP)-expressing cells inside the liver was similar in wild-type and mutants (Fig. 5E), providing direct evidence that HSCs do not derive from endothelial cells and that they do not rely on SECs to enter the liver.
In mammals, HSCs are situated in the space of Disse, which sits beneath the basolateral surface of hepatocytes, whereas biliary cells are restricted to the portal regions (reviewed by 27). We found that in wild-type zebrafish livers, the cell bodies of HSCs are separated from biliary cells by hepatocytes (Fig. 5C). Strikingly, in clo mutant livers, HSCs appear to be closely associated with biliary cells (Fig. 5D, arrows), suggesting that loss of SECs alters the relationships amongst HSCs, hepatocytes and biliary cells.
The angiogenic factor VEGF and its receptors are induced in activated HSCs during liver injury, and VEGF signaling has been shown to mediate the cross talk between SECs and HSCs (reviewed by 1). To determine whether VEGF is also involved in SEC-HSC interactions during development, we treated Tg(hand2:EGFP;kdrl:ras-mCherry) animals with SU5416, a potent and selective inhibitor of the Flk-1/KDR receptor tyrosine kinase (28). SU5416 has been shown to effectively block VEGF signaling and angiogenesis in zebrafish (29, 30). In agreement with these reports, we observed a dose-dependent decrease in the number of intrahepatic vascular branches in animals treated with SU5416 between 55 and 80 hpf (Fig. 6A-B). The number of HSCs was significantly reduced upon SU5416 treatment (Fig. 6A, C), although most remaining HSCs still kept close contact with SECs (data not shown). When we treated animals with SU5416 between 72 and 96 hpf, after significant numbers of SECs and HSCs had entered the liver, we observed a much milder deficiency of SECs and HSCs (Fig. 6A-C).
To test the specific requirement of Kdrl in HSC and SEC development, we injected Tg(hand2:EGFP; kdrl:ras-mCherry) embryos with an antisense morpholino (MO) targeted against kdrl (18). Consistent with previous data (18), injection of 3 ng of kdrl MO resulted in a severe reduction of blood vessels (data not shown). We did not detect any Tg(kdrl:ras-mCherry) expression in the liver in kdrl-knock down animals (Fig. 6D). Similar to the phenotypes caused by SU5416 treatment, we observed a significant reduction of HSCs in kdrl-knock down animals (Fig. 6D). Such a reduction was not likely due to a delay in liver growth as the numbers of Prox1+ parenchymal cells were similar between uninjected controls and kdrl-knock down animals (p>0.11, 10 control and 10 kdrl-knock down animals were analyzed).
The modulation of VEGF signaling may directly affect both SECs and HSCs. However, it is also possible that SECs are primarily affected, with the ensuing decrease in HSCs resulting from alterations in signaling between SECs and HSCs. If the latter scenario was correct, inhibition of VEGF signaling should not affect HSCs in the absence of SECs. We therefore conducted SU5416 treatments on clo mutants and found that they caused a dose- and stage-dependent decrease in HSC cell number similar to the decrease seen in wild-type (Fig. 6E). This result indicates that inhibition of VEGF signaling independently impairs both HSCs and SECs.
The identification of bioactive compounds that affect HSC behavior could provide insight into HSC biology and offer means to manipulate HSCs during liver injury or disease. We performed a pilot screen of 338 compounds, looking for agents that caused either a decrease or an increase in HSC numbers. Animals were treated with compounds from 55 to 80 hpf (Fig. 7A; Supporting Information). We found that AM580, a retinoic acid receptor (RAR)-alpha-selective agonist (31), decreased HSC numbers (Fig. 7B-C). By contrast, methoprene acid (MA), a retinoid X receptor (RXR) agonist (32), increased the numbers of HSCs (Fig. 7B-C). Neither drug affected gross liver morphology or larval survival. These data indicate that retinoid receptor agonists alter HSC numbers in vivo, which is consistent with a previous report that retinoic acid signaling regulates HSC development (33). Furthermore, the identification of two HSC-altering compounds in our pilot screen validates this approach as a way to discover compounds that impact HSC biology.
Here we present the first HSC reporter line in zebrafish, Tg(hand2:EGFP). Similar to mammalian HSCs, zebrafish HSCs are in close proximity to SECs, exhibit stellate morphology and store lipid droplets. Importantly, they display robust cellular responses to acute ethanol exposure, and thus are functionally equivalent to their mammalian counterparts in response to this hepatic insult. We tracked the development of HSCs and show that they enter the liver after SECs do. By analyzing clo mutants, we reveal that SECs are not required for HSC differentiation or their entry into the liver. However, in the absence of SECs, HSCs associate instead with biliary cells. We also show that inhibition of VEGF signaling significantly reduces the number of HSCs that enter the liver, even in the absence of SECs. Lastly, we demonstrate the use of the HSC reporter line in gene-profiling analyses and chemical screens to identify novel molecular mechanisms underlying HSC development. Our study provides a detailed characterization of HSC development in zebrafish, and validates the usefulness of this model organism in HSC research.
The embryonic origin of HSCs is elusive due to their diverse gene expression. We show that in zebrafish, HSCs express hand2 which labels mesodermal and neural crest derivatives and supports a mesodermal and neural crest origin for HSCs. It has been hypothesized that HSCs and SECs share a common precursor. Studies of HSC-SEC interactions have been limited, in part because mouse mutants deficient in endothelial cells exhibit early embryonic lethality (34). Unlike the mammalian liver, the zebrafish liver is not a hematopoietic organ and liver defects do not lead to anemia or early lethality (25). We detected HSCs in clo mutants which lack SECs, thereby providing direct evidence that HSCs do not originate from SECs or their precursors in zebrafish.
Our analyses of clo mutants reveal a surprising role for SECs in the localization of HSCs during liver development. In normal livers, SECs guide HSCs to their proper sinusoidal location. When SECs are absent, HSCs do not fail to migrate into the liver; instead, they become abnormally associated with biliary cells. It is possible that in normal livers, SECs and biliary cells both have the capacity to control HSC migration, with SECs being dominant. Alternatively, biliary cells may have no inherent ability to attract HSC in normal livers, but in the absence of SECs, may acquire a novel phenotype with this capacity. In biliary diseases, cholangiocytes are thought to undergo epithelial-to-mesenchymal transition (EMT), by which they acquire features of mesenchymal cells (reviewed by 35). It will be interesting to determine whether EMT of biliary cells contributes to the altered HSC distribution in clo mutants. Biliary-HSC interactions play important roles in biliary diseases, and paracrine signals including chemokines, cytokines, purinergic agonists, and morphogens such as Hedgehog have been shown to mediate these interactions (reviewed by 35). Further study of clo mutants may identify additional paracrine signals, thereby advancing our understanding of biliary-HSC interactions in normal and diseased livers.
We show that VEGF signaling inhibition via chemical inhibitor treatments or kdrl MO injections decreases HSC numbers and that this effect still occurs in the absence of SECs. VEGF signaling does not seem to be essential for differentiation and survival of HSCs, as blocking VEGF signaling during later stages only causes a mild decrease in HSC numbers. Rather, VEGF signaling may be required for the initial wave of HSCs entering the liver. During later development, when significant numbers of HSCs and SECs are present in the liver, different signals, possibly from the existing HSCs and SECs, might play a greater role in attracting additional HSCs. Meanwhile, because both of our approaches induced global inhibition of VEGF signaling, it is plausible that they also impair hepatocytes and/or biliary cells, which may contribute to the reduction in HSC numbers. In the future, it will be necessary to knock down VEGF receptor function selectively in HSCs to determine the HSC-specific requirement of VEGF signaling.
We describe a pilot chemical screen to identify drugs affecting HSC development. Chemical screens using zebrafish have several advantages over those performed in cultured cells (36, 37). First, the zebrafish system enables one to identify drugs that cause specific phenotypes in HSCs without causing substantial toxicity to other organs. Second, in vivo systems enable the identification of compounds that require the presence of additional cell types, matrix components and/or growth factors to help mediate their effects. Finally, drugs identified through in vitro screens may not be effective in live animals, perhaps because of critical differences in gene expression (37). Our pilot screen resulted in the identification of two retinoid receptor agonists, AM580 and MA, that caused opposite effects on HSC numbers. Although HSCs have a well-established role in retinoid storage and transport, the role of retinoids in regulating HSC proliferation and fibrogenesis is incompletely understood, and reports investigating the effects of retinoids on HSCs have been contradictory (reviewed by 1). Interestingly, all-trans-retinoic acid (RA), 9-cis-RA, and synthetic retinoids have divergent effects on activated HSCs in culture (38), supporting the hypothesis that RXR and RAR have at least some non-overlapping effects in HSCs. It will be interesting to further explore the role of retinoid signaling during development and injury using the Tg(hand2:EGFP) line, as well as to extend the chemical screen to identify additional compounds with effects on HSC biology.
Our work demonstrates that the Tg(hand2:EGFP) line represents a versatile in vivo model for HSC research that complements cell culture and mammalian systems. More importantly, it illustrates that this HSC reporter line can be applied not only to the study of HSC development, but also to the study of their activation in liver injury, as acute alcohol insult triggers robust cellular responses of zebrafish HSCs similar to those observed in activated HSCs in mammalian liver injury. The functional similarities between zebrafish and mammalian HSCs, combined with the unique strengths of the zebrafish model, open up numerous exciting new avenues for studying HSC biology in liver injury.
We would like to thank Drs. Scott Friedman and D. Montgomery Bissell for critical comments and support, and Stainier lab members for technical advice and discussions. We acknowledge the UCSF Liver Center for technical support and the UCSF Genomics Core Facility for conducting quantitative real-time PCR. We thank Sarah Elmes for assistance with FACS and Ana Ayala, Milagritos Alva and Mark Sklar for fish care.
Chunyue Yin is supported by Grant Number K99AA020514 from the NIH, and the UCSF Liver Center Pilot/Feasibility Award (NIH P30DK026743); Kimberley J. Evason is a Damon Runyon Fellow supported by the Damon Runyon Cancer Research Foundation (DRG-109-10); Jacquelyn J. Maher is supported by grants from the NIH (R01DK068450, R01DK088674, and P30DK026743); this work was supported in part by grants from the NIH (R01DK060322) and the Packard Foundation to Didier Y. R. Stainier.