Potential conflict of interest: Nothing to report.
Liver regeneration is impaired following partial hepatectomy (PH) in mice with genetic obesity and hepatic steatosis and also in wild-type mice fed a high-fat diet. These findings contrast with other data showing that liver regeneration is impaired in mice in which hepatic lipid accumulation is suppressed by either pharmacologic leptin administration or by disrupted glucocorticoid signaling. These latter findings suggest that hepatic steatosis may actually be required for normal liver regeneration. We have reexamined this relationship using several murine models of altered hepatic lipid metabolism. Liver fatty acid (FA) binding protein knockout mice manifested reduced hepatic triglyceride (TG) content compared to controls, with no effect on liver regeneration or hepatocyte proliferation. Examination of early adipogenic messenger RNAs revealed comparable induction in liver from both genotypes despite reduced hepatic steatosis. Following PH, hepatic TG was reduced in intestine-specific microsomal TG transfer protein deleter mice, which fail to absorb dietary fat, increased in peroxisome proliferator activated receptor alpha knockout mice, which exhibit defective FA oxidation, and unchanged (from wild-type mice) in liver-specific FA synthase knockout mice in which endogenous hepatic FA synthesis is impaired. Hepatic TG increased in the regenerating liver in all models, even in animals in which lipid accumulation is genetically constrained. However, in no model—and over a >90-fold range of hepatic TG content—was liver regeneration significantly impaired following PH. Conclusion: Although hepatic TG content is widely variable and increases during liver regeneration, alterations in neither exogenous or endogenous lipid metabolic pathways, demonstrated to promote or diminish hepatic steatosis, influence hepatocyte proliferation. (HEPATOLOGY 2008.)
The mammalian liver has a remarkable capacity to regenerate. For example, following experimental partial hepatectomy (PH), with removal of 70% of the liver, hepatocytes begin within hours to respond to signals that initiate and sustain a complex yet well-integrated process of proliferation.1, 2 Growth factor–dependent priming of liver regeneration promotes the reentry of quiescent hepatocytes into the cell cycle, while progression through the restriction point in late growth phase 1 (G1) is maintained by other growth factors, cyclins, and their respective kinases, perpetuating a proliferative drive that over several days restores hepatocyte mass (reviewed in Fausto et al.2). However, despite intensive study over the last several years, intersections between these molecular signaling cascades and the multiple metabolic pathways that may influence liver regeneration remain incompletely characterized.
Integral to the successful initiation and completion of liver regeneration is that the remaining cells within the liver acquire sufficient energy substrate to support the metabolic demands of rapid proliferation. It has long been recognized that the regenerating liver generates signals (still poorly understood) that couple fatty acid (FA) release from peripheral adipose stores to augmented hepatic FA uptake, which in turn promotes hepatic lipogenesis3 and leads to rapid accumulation of intracellular triglyceride (TG) within the regenerating liver4 (reviewed in Brasaemle,5 Farrell,6 and Rudnick7). However, despite the observation that liver regeneration is physiologically associated with transient steatosis following PH, there is considerable evidence that liver regeneration is impaired in certain genetic models in which the liver contains excess fat. For example, both ob/ob mice and db/db mice exhibited up to 70% mortality following PH, with defective liver regeneration in the surviving animals.8–10 One interpretation of these findings is that impaired leptin signaling is responsible for defective liver regeneration in these models,6 although this conclusion merits reconsideration in light of more recent work demonstrating that leptin repletion failed to reverse this phenotype in ob/ob mice.11 Other studies reveal that hepatic steatosis induced by high-fat feeding in C57BL/6 mice resulted in impaired liver regeneration following PH, suggesting that hepatic steatosis per se rather than impaired leptin signaling may be a key element, although the possibility that leptin signaling plays a role in this particular model could not be discounted.12 By contrast, other nutritional models of steatohepatitis, for example feeding a methionine-choline deficient diet, failed to alter liver regeneration in rats following PH.13
Compounding the complexity in some of these apparently conflicting findings, there is also evidence that genetic or pharmacologic approaches that reduce lipid accumulation may also impede liver regeneration. In one such study, Cav-1−/− mice subjected to PH failed to accumulate hepatic TG during liver regeneration and exhibited decreased proliferation and ∼70% mortality between 48 and 72 hours.14 However, another study using a different genetic strain of Cav-1−/− mice revealed no increase in mortality following PH and demonstrated that wild-type (WT) and Cav-1−/− mice accumulated comparable TG during liver regeneration.15 In yet another study, mice injected with pharmacologic doses of leptin prior to PH manifested a 50% reduction in hepatic TG content and a marked suppression of hepatocyte proliferation following PH.16 These studies also demonstrated that liver-specific knockout of the glucocorticoid receptor was associated with both a 50% reduction in hepatic TG and also defective regeneration following PH.16 On balance, these studies collectively raise the possibility that hepatic steatosis may actually represent a required component for liver regeneration following PH and that maneuvers which diminish hepatic steatosis may also suppress liver regeneration.5, 6
In the present study, we have addressed certain unresolved findings emerging from these various reports. In particular, we wished to address the hypothesis that the capacity of the liver to regenerate following PH was impaired in a setting in which hepatic steatosis was constrained, but without alterations in either leptin or glucocorticoid signaling pathways and without dietary manipulation. We selected liver FA binding protein knockout (L-Fabp−/−) mice because they display reduced rates of FA uptake and utilization in the liver and are protected against the development of both hepatic steatosis and diet-induced obesity when fed a high saturated FA diet.17, 18 In addition, we studied mice with defective intestinal lipid absorption and delivery, mice with defective hepatic FA oxidation, and mice with a liver-specific defect in FA synthesis to obtain a range of models in which hepatic fat metabolism was impaired in circumstances under which the animals consumed a chow diet.19–21 The summary conclusion of our study is that over a >90-fold range of hepatic TG content there was no detectable effect of these genetic changes on liver regeneration following PH.
Biochemical assays for TG, cholesterol, free fatty acid (FFA), phospholipids (PLs), glucose, and β-hydroxybutyrate levels in serum or tissue were performed using kits obtained from Wako Chemicals (Richmond, VA). Serum alanine aminotransferase levels were determined using a kit from Teco Diagnostics (Anaheim, CA).
L-Fabp−/− mice are C57BL/6 congenic.17 Microsomal TG transfer protein (MTP) intestinal knockout (IKO) model (MTP-IKO) mice were generated as described19 (also see the Supplementary Methods). Peroxisome proliferator-activated receptor (PPAR)α−/− mice are C57BL/6J congenic,20 and were crossed with L-Fabp−/− mice to produce PPARα−/− L-Fabp−/− double knockout (PLDKO) mice. Mice with liver specific disruption of fatty acid synthase (FAS) (FAS–knockout in liver [KOL]; that is, “FAS-KOL” mice) were generated as described21 (also see the Supplementary Methods). All animal protocols were approved by the Washington University Animal Studies Committee and conformed to criteria outlined in the National Institutes of Health “Guide for the Care and Use of Laboratory Animals.”
Animals were 10-20 weeks of age. A standard 70% hepatectomy was performed in which the left and median lobes of the liver were removed. Resected liver tissue was weighed and frozen in liquid nitrogen for later analysis (0 hour liver). Micro-osmotic pumps (Alzet; Model 1003D) containing 100 μl of bromodeoxyuridine (BrdUrd; 18 mg/ml) and fluorodeoxyuridine (1.8 mg/ml) were implanted to measure hepatocyte proliferation. At sacrifice, serum and tissues were collected and stored for later analysis. The surgery survival rate was >95%, with no difference between the genotypes. Additional details are provided in the Supplementary Methods.
Analysis of Tissue Lipid Content.
Frozen liver tissue (∼100 mg) was homogenized in 1.6 ml phosphate-buffered saline and protein concentration was determined using DC Protein Assay (Bio-Rad, Hercules, CA). Lipid was extracted using chloroform:methanol (2:1) and 0.1% sulfuric acid as described.22 An aliquot of the organic phase was collected, dried with chloroform containing 1% Triton, and resuspended in water (final Triton concentration = 2%). TG, FFA, PL, and cholesterol content were determined using commercially available kits (Wako Chemicals) in microtiter plates and normalized to protein concentration of the homogenate.
Liver tissue was fixed in 10% buffered formalin and embedded in paraffin. BrdUrd (also BrdU) incorporation was assessed by immunostaining with anti-BrdUrd immunoglobulin G (Accurate Chemical; H8365) followed by secondary staining with anti-rat immunoglobulin G (Jackson Laboratories). Sections were counterstained with hematoxylin. For quantitation of BrdU incorporation, BrdU-positive and BrdU-negative nuclei were counted in at least four high-power fields per animal and expressed as the percentage of BrdU-positive nuclei. Ki67 antigen staining was performed on paraffin sections using a rat anti-mouse Ki67 monoclonal antibody (Clone TEC-3; Dako North America, Carpinteria, CA), counterstained with hematoxylin, and quantitated as described above.
RNA was isolated from frozen liver tissue using Trizol (Invitrogen, Carlsbad, CA). Total RNA (10 μg) was treated with deoxyribonuclease (DNase) I (DNA-Free; Ambion) and used for the production of complementary DNA using SuperScript II RT (Invitrogen) as described.18 Real-time quantitative PCR was performed on an SDS 7000 (Applied Biosystems) using 2× SYBR Green Master Mix (Applied Biosystems). Expression levels were normalized to the 18S gene, and relative gene expression was determined using the comparative threshold cycle method (Applied Biosystems; User Bulletin 1). Oligonucleotide primers for quantitative PCR are listed in the Supplementary Methods.
Antisense Oligonucleotide Treatment.
Antisense oligonucleotides (ASO) were designed and synthesized by ISIS Pharmaceuticals, Inc. (Carlsbad, CA). Male C57BL/6 mice (Jackson Laboratory) were injected intraperitoneally with 50 mg/kg of control or L-Fabp ASO twice a week for 4-6 weeks. Additional details are provided in the Supplementary Methods.
Statistical significance was determined using an unpaired, two-tailed Student t test, performed using Microsoft Excel. Trend line and correlation coefficient (Fig. 5) were determined using Microsoft Excel. Unless otherwise noted, data are expressed as mean ± standard error.
L-Fabp−/− Mice Exhibit Reduced Hepatic Steatosis Yet Normal Regeneration Following PH.
Hepatic TG content was elevated in WT mice with peak (∼10-fold elevated) steatosis occurring 12-24 hours after PH (Fig. 1A), as demonstrated.4 In contrast, while the overall trend was maintained, the magnitude of PH-associated steatosis was significantly blunted in L-Fabp−/− mice at all times following PH (Fig. 1A). Changes in hepatic cholesterol, FFA, and PL content following PH were not as dramatic, and no differences were observed between the genotypes (data not shown). Serum TG decreased (Fig. 1B) while serum FA levels increased dramatically (Fig. 1C) in both genotypes immediately after PH, indicating equivalent release of FA from adipose tissue deposits. In contrast, serum β-hydroxybutyrate levels were reduced in L-Fabp−/− mice, both basally and following PH (Fig. 1D), consistent with reduced uptake and availability of hepatic FA for ketogenesis.17, 18 Serum alanine aminotransferase activity increased dramatically 6 hours after PH and declined thereafter, with no difference between the genotypes (Supplementary Fig. 1).
In view of findings demonstrating that a two-fold reduction in hepatic TG content at 48 hours following PH was associated with reduced proliferation following PH,16 we next examined whether the recovery of liver mass was impaired in L-Fabp−/− mice. Survival following PH was virtually quantitative in both genotypes at all times (data not shown) and liver weight (expressed as % of total body weight at sacrifice) was generally comparable between the genotypes (Fig. 2A). There was also no significant difference in hepatocyte proliferation between WT and L-Fabp−/− mice at 48 or 72 hours after PH, as determined either by cumulative BrdU incorporation (Fig. 2B; Supplementary Fig. 2A) or by expression of the cell proliferation marker Ki67 (Fig. 2C; Supplementary Fig. 2B). In addition, similar albeit variable induction of the cell cycle genes cyclin D1 and p21 was observed in the livers of both genotypes 24 hours after PH (1.3 ± 0.2-fold and 3.4 ±1.3-fold induction of cyclin D1; 56.4 ± 28.7-fold and 46.1 ± 26.5-fold induction of p21, in C57BL/6 and L-Fabp−/− animals, respectively, compared to expression in 0 hour livers), with levels of induction similar to those observed in other studies.12, 23 Together, these data indicate that liver regeneration is not significantly impaired in L-Fabp−/− mice, despite a striking decrease in PH-associated steatosis.
Normal Regeneration Following Liver-Specific Knockdown of L-Fabp with Reduced Hepatic Steatosis Following PH.
C57BL/6 mice treated with L-Fabp ASO displayed a ∼60% reduction in L-Fabp messenger RNA (mRNA) and protein (Supplementary Fig. 3) in the liver compared to mice treated with control ASO, with no change in intestinal L-Fabp mRNA. Following PH, hepatic TG content was reduced ∼50% in the livers of the L-Fabp ASO-treated mice compared to control ASO-treated mice at 24 hours after PH, though no difference was observed at 48 hours and there was no difference in the percentage of BrdU-positive nuclei between the two treatment groups (Supplementary Fig. 3). These findings strongly implicate the decrease in hepatic L-Fabp gene expression as an underlying factor in abrogating the extent of hepatic TG accumulation following PH, but again suggest that decreased hepatic FA availability and utilization per se do not impair regeneration.
Decreased Intestinal TG Absorption and Reduced Dietary FA Delivery Does Not Impair Liver Regeneration Following PH.
To examine another possible genetic pathway whereby reduced hepatic steatosis might impair liver regeneration following PH, we turned to the intestine-specific MTP-IKO model, in which dietary TG absorption is virtually eliminated, and the delivery of dietary FA is severely restricted. Hepatic lipid content was reduced in MTP-IKO mice at 24 and 48 hours after PH, with a pattern indistinguishable from that observed in L-Fabp−/− mice (Fig. 3A). In addition, and despite the virtual absence of visible peripheral adipose tissue deposits in the MTP-IKO mice (data not shown),19 serum FFA levels were higher in these mice than in C57BL/6 controls 6 hours after PH but fell by more than 50% at 24 hours (Fig. 3B). However, liver regeneration was not decreased in these mice 48 hours after PH (% BrdU-positive cells: C57BL/6, 39.2 ± 6.2; MTP-IKO, 36.1 ± 3.3; n = 5-9), nor was there a defect in the induction of cyclin D1 and p21 24 hours after PH (data not shown), suggesting again that decreased hepatic TG accumulation, in this case the result of severely restricted dietary FA delivery rather than defective hepatic FA uptake, does not impair liver regeneration. Note that hepatic MTP mRNA levels were similarly induced in C57BL/6 and MTP-IKO mice following PH (Fig. 3C), an observation we will return to below (also discussed in the Supplementary Results and Supplementary Fig. 4B).
Hepatic Adipogenic Gene Program Is Induced Despite Abrogation of Hepatic Steatosis.
Previous studies documented a rapid and dramatic upregulation of adipogenic and other adipose-specific genes, including adipsin, Fabp4, and Fsp27, in the livers of WT mice following PH, concomitant with or preceding the increase in lipid accumulation.16, 24 In view of the divergence in hepatic TG content between WT and L-Fabp−/− mice following PH, we examined whether these changes might reflect corresponding alterations in the expression of these adipogenic genes using real-time quantitative PCR. However, this was not the case. The induction of Fsp27, S3-12, Adipsin, Adrp (adipophilin), and Fabp4 was similar in the livers of both genotypes at 6-12 hours, despite the divergence in hepatic TG content (Supplementary Fig. 4A). Overall, these data indicate that the induction of “adipogenic” genes in the liver is not impaired in L-Fabp−/− mice, and thus cannot explain the reduced TG accumulation observed in these animals. Moreover, it appears that induction of this adipogenic program is not quantitatively driven in relation to the accumulation of lipid.
Defective FA Oxidation in PPARα −/− Mice Leads to Augmented Hepatic Accumulation But Does Not Impair Liver Regeneration Following PH.
There are conflicting findings in regard to liver regeneration following PH in PPARα−/− mice, with some studies suggesting delayed regeneration while others demonstrating no change.25–27 Some of this apparent discrepancy may reflect subtle differences in the genetic backgrounds of the various lines, with some studies performed in 129/SvJ mice while others were performed in a partially backcrossed C57BL/6 background. To extend our comparisons of liver regeneration following PH in genetic models with altered hepatic TG accumulation, we examined PPARα−/− mice as well as PLDKO mice and their congenic C57BL/6 controls. Following PH, hepatic TG content was increased and remained elevated in PPARα−/− mice 48 hours after PH (Fig. 4A), in association with an inability to upregulate expression of MTP mRNA (Fig. 4B). The lipid accumulation phenotype of PPARα−/− mice at 48 hours was reversed in the PLDKO mice (Fig. 4A), in conjunction with a corresponding increase in MTP mRNA expression, suggesting that the regulation of MTP gene expression is at least partially independent of PPARα (Fig. 4B). Nevertheless, despite the range of hepatic TG accumulation noted in this series of experiments, hepatic regeneration as evidenced by BrdU incorporation or induction of cell cycle genes was indistinguishable among the various genotypes (PLDKO, 37.8 ± 7.8% BrdU-positive; PPARα−/−, 28.6 ± 4.6% BrdU-positive; C57BL/6, 39.2 ± 6.2% BrdU-positive; n = 7-9 per genotype, and data not shown).
In addition to the models detailed above, we also examined liver regeneration in mice with liver-specific deletion of FAS (FAS-KOL mice) since these animals have a defect in endogenous hepatic lipogenesis that might influence their capacity to sustain liver regeneration under conditions of metabolic stress.21 However, yet again our observations failed to bear out this possibility. FAS-KOL mice demonstrated an indistinguishable pattern of transient hepatic steatosis and liver regeneration compared to WT controls (Fig. 4C) and no difference in BrdU incorporation (48.0 ± 3.6% BrdU-positive nuclei versus 39.2 ± 6.2% BrdU-positive nuclei; FAS-KOL versus C57BL/6, respectively; n = 6-9; P = 0.3).
Hepatic TG Accumulation Does Not Predict the Extent of Hepatocyte Proliferation Following PH.
We have evaluated the relationship between hepatic lipid content at 48 hours and liver regeneration as inferred by cumulative BrdU labeling over the same period in multiple lines of mice with altered hepatic lipid metabolism. Overall, it appears that there is no correlation between the extent of hepatic steatosis—over a >90-fold range of hepatic TG content—and hepatocyte proliferation. These data are shown graphically in Fig. 5, illustrating the range of hepatic TG accumulation in individual animals among the various models. Despite significant variability in the percentage of BrdU-positive cells, there is clearly no overall correlation between the number of replicating cells and the amount of TG within the liver. For example, the two data points highlighted in Fig. 5 display a 30-fold difference in hepatic TG content, with comparable hepatocyte proliferation.
One of the major hallmarks of liver regeneration following injury or PH is the striking yet transient steatosis that occurs from ∼12 to 48 hours, with a peak prior to the initial wave of hepatocyte proliferation.4 The mechanisms underlying this accumulation of TG include increased FA mobilization and delivery to the liver, coupled with increased hepatic lipogenesis and decreased secretion of very low density lipoprotein.3, 4 Recent studies have also demonstrated a temporal induction of adipogenic genes in hepatocytes prior to the onset of hepatic steatosis, suggesting that this may represent a programmed adaptation required for the regenerative response.16, 24 The central observation to emerge from the current study is that the transient hepatic steatosis that accompanies liver regeneration following PH may vary nearly 100-fold, yet its magnitude alone does not predict hepatocyte proliferation. This observation must be considered against the background of previous studies that have linked these two events—but with contrasting implications. On the one hand, some studies demonstrate that genetic or dietary manipulations that promote hepatic lipid accumulation also reduce liver regeneration. By contrast, other studies demonstrate that genetic or pharmacological maneuvers which decrease hepatic steatosis may actually impair the proliferative response.
Earlier studies performed with either ob/ob,8, 10 or db/db mice9 demonstrated that leptin signaling is required for normal liver regeneration, and raised the possibility that the associated metabolic phenotype of obesity, hepatic steatosis, and insulin resistance might account for at least some of the observed defect. This explanation, however, appears inconsistent with more recent observations demonstrating that exogenous leptin supplementation (3 weeks) of ob/ob mice, a regimen sufficient to reverse the obesity and metabolic phenotype (with or without normal circulating leptin levels), failed to reverse the defect in hepatocyte proliferation following PH.11 Accordingly, it would be reasonable to conclude that hepatic steatosis per se appears unrelated to the impairment of liver regeneration, even in the background of leptin deficiency.
We further examined this association using PPARα−/− mice in a congenic C57BL/6 background and again find no defect in hepatocyte proliferation or liver regeneration. Our findings demonstrate increased accumulation of hepatic TG following PH in this genotype26 (Fig. 4A) and suggest that a contributing mechanism, in addition to the defects in FA oxidation, may be impaired upregulation of MTP expression, a known PPARα target.28 The finding that augmented hepatic TG accumulation can be un-coupled from defective hepatic proliferation following PH is in agreement with studies in acetyl coenzyme A oxidase knockout mice, in which there was no alteration in hepatic regeneration following PH despite extensive microvesicular steatosis.29
We also examined murine genetic models in which hepatic steatosis was impaired following PH, to address the hypothesis that interruption of metabolic pathways which promote hepatic steatosis might actually constrain liver regeneration. This hypothesis emerged from studies in WT mice infused with pharmacologic doses of leptin prior to PH in which liver regeneration was impaired in conjunction with decreased hepatic TG accumulation, an observation subsequently confirmed by others.11, 16 In addition, mice with liver-specific deletion of the glucocorticoid receptor demonstrated both impaired liver regeneration and also demonstrated an approximately two-fold decrease in hepatic TG content.16 In one approach, we used L-Fabp−/− mice since these animals display impaired FA uptake, decreased very low density lipoprotein production, and reduced hepatic TG following a 48 hour fast, a setting in which adipose tissue TG stores are mobilized for FA delivery.17 In addition, previous studies have demonstrated that L-Fabp gene expression is highly induced following PH, suggesting that this gene may be required for the metabolic adaptations that take place during liver regeneration.30 Our studies reveal decreased hepatic TG accumulation following PH in L-Fabp−/− mice, with no impairment of liver regeneration.
A previous study in Cav-1−/− mice demonstrated defective liver regeneration in association with a failure to accumulate hepatic TG.14 Although not confirmed in a subsequent study,15 these data raised the possibility that perhaps selected metabolic defects which lead to impaired hepatic lipid accumulation might similarly impact liver regeneration. As an alternative approach to test this hypothesis, we used MTP-IKO mice in which there is virtually no dietary fat absorption and there is compensatory upregulation of hepatic lipogenesis in the setting of reduced hepatic TG.19 However, despite the predicted impairment of hepatic TG accumulation in this model, there was again no detectable effect on hepatocyte proliferation or liver regeneration.
It bears emphasis that the reductions observed in hepatic TG after PH in MTP-IKO and L-Fabp−/− mice equal or exceed that reported following pharmacologic leptin administration or abrogated glucocorticoid signaling (generally approximately two-fold),16 suggesting that our failure to impact liver regeneration cannot be attributed simply to an inadequate reduction in steatosis. It is worth noting that hepatic TG content indeed increased in L-Fabp−/− mice during liver regeneration—albeit less dramatically than observed in WT controls—preceded by induction of adipogenic mRNAs in the regenerating liver, with a temporal pattern similar to that from WT controls. We infer from these results that diminished hepatic TG accumulation following PH—independent of hormonal signaling pathways—does not appear to decrease hepatocyte proliferation or liver regeneration. Our findings do, however, raise the intriguing possibility that even in circumstances under which TG accumulation is genetically constrained, there is a threshold of adaptive lipogenesis that we have yet to influence. This possibility will require further study.
It is worth recognizing that other murine genetic models, for example Nrf2−/− mice, manifest defective liver regeneration accompanied by augmented hepatic steatosis.31 However, Nrf2−/− mice exhibit additional defects, including increased oxidative stress and defective insulin and insulin-like growth factor 1 signaling, which more likely contribute to the phenotype observed following PH, rather than hepatic steatosis per se.31 These caveats, along with the reservations noted above in regard to leptin signaling, limit conclusions that directly link increased hepatic steatosis to defective liver regeneration. The finding that hepatic steatosis in transgenic diacylglycerol acyltransferase 2 mice is not accompanied by insulin resistance is strong evidence suggesting that the metabolic consequences of hepatic TG accumulation may indeed be uncoupled from other important functions.32 Among the possibilities in this regard are that defective liver regeneration in the setting of steatosis results from alterations in other mediators such as inflammatory or hormonal signals. Our findings demonstrating that hepatic TG accumulation—in a range of murine genetic models with altered lipid metabolism—may be uncoupled from defective liver regeneration following PH suggest that fat alone is not the culprit.
We acknowledge the support of the Digestive Disease Research Core Center (DDRC) Morphology Core at Washington University School of Medicine. In addition, the authors are grateful to Kelly Waters for technical assistance and to Dr. David Rudnick and Dr. Valerie Blanc for informative discussions.