Metabolic and non-metabolic liver zonation is established non-synchronously and requires sinusoidal Wnts

The distribution of complementary metabolic functions in hepatocytes along a portocentral axis is called liver zonation. Endothelial secreted Wnt ligands maintain metabolic zonation in the adult murine liver but whether those ligands are necessary to initiate zonation in the immature liver has been only partially explored. Also, numerous non-metabolic proteins display zonated expression in the adult liver but it is not entirely clear if their localization requires endothelial Wnts. Here we used a novel transgenic mouse model to compare the spatial distribution of zonated non-metabolic proteins with that of typical zonated metabolic enzymes during liver maturation and after acute injury induced by carbon tetrachloride (CCl4). We also investigated how preventing Wnt ligand secretion from endothelial cells affects zonation patterns under homeostasis and after acute injury. Our study demonstrates that metabolic and non-metabolic zonation are established non-synchronously during maturation and regeneration and require multiple endothelial Wnt sources.


Introduction
The liver executes multiple functions involved with anabolic and catabolic metabolism, bile and protein secretion, drug detoxification, glycogen storage, and other processes. These tasks are performed simultaneously and with maximal energy-efficiency thanks to a unique anatomical architecture consisting of hexagonal functional units (a.k.a. 'lobules'), typically arranged in a honeycomb-like pattern (Gebhardt and Matz-Soja, 2014). At the periphery of the lobules, blood enters the portal vein and the hepatic artery and flows into the liver sinusoids towards the central vein. Inside the lobules, the highly polarized hepatocytes assemble into cords that face the hepatic sinusoids on the basolateral side and form canaliculi on the apical side (Tanaka and Miyajima, 2016). Although morphologically quite similar, hepatocytes accomplish very different metabolic roles according to their specific position along the lobular portocentral axis. The segregation of hepatocytes into discrete functional areas is referred to as 'Liver Zonation' and is the basis for the partition of the hepatic lobule into three metabolic zones: periportal Zone 1 consisting of 6-8 hepatocyte layers that receive blood enriched in oxygen and nutrients and control glycogen metabolism, amino acid utilization and ammonia detoxification; intermediate perivenous Zone 2 consisting of 6-10 hepatocyte layers with a major role in xenobiotic metabolism; and perivenous Zone 3 formed by 2-3 hepatocyte layers that surround the central veins and perform biotransformation reactions, glutamine synthesis and glycolysis (Jungermann and Katz, 1989). Metabolic zonation optimizes liver physiology by quickly adapting this organ to exogenous nutritional challenges and endogenous metabolic demands. Moreover, pathologic alterations that permanently change zonation patterns lead to metabolic disease (Gebhardt and Matz-Soja, 2014).
Pioneer studies revealed that canonical Wnt/b-catenin signaling preserves metabolic zonation in the adult liver (Benhamouche et al., 2006;Burke et al., 2009;Planas-Paz et al., 2016;Sekine et al., 2006;Yang et al., 2014) and promotes the expression of pericentral enzymes involved with glutamine synthesis in the late fetal liver (Burke et al., 2018). More recent investigations also demonstrated that Wnt ligands and the Wnt agonist R-spondin3 produced by hepatic endothelial cells are necessary to preserve metabolic zonation in the adult liver (Leibing et al., 2018;Planas-Paz et al., 2016;Preziosi et al., 2018;Rocha et al., 2015;Wang et al., 2015). While these and other studies have increased our knowledge of adult metabolic zonation, we still do not know much about how metabolic and non-metabolic proteins begin to segregate into specific regions in the maturing liver or how zonated protein expression is reestablished in hepatocytes following acute hepatic injury. Moreover, it has been shown that hepatic sinusoidal cells produce Wnt ligands (Ding et al., 2010;Wang et al., 2015) but the potential contribution of this specialized endothelium to liver zonation recovery after acute injury has not been directly proven.
Here we used a novel transgenic mouse strain (Cldn2-EGFP)  to investigate how the tight junction protein claudin-2 becomes zonated in hepatocytes during liver maturation and after CCl 4 -induced acute injury. We also used conditional deletion approaches and Cldn2-EGFP mice to investigate the role of hepatic sinusoid-derived Wnt ligands in the establishment and maintenance of liver zonation, and the role of endothelial-derived Wnt ligands in zonation restoration after CCl 4 -induced damage.

Spatiotemporal analysis of postnatal liver zonation using Cldn2-EGFP transgenic mice
Published studies showed that the tight junction protein claudin-2 exhibits zonated expression in perivenous areas of the adult murine liver (Matsumoto et al., 2014;Rahner et al., 2001). Using immunohistochemistry analysis, we corroborated claudin-2 expression almost exclusively in perivenous areas in the adult mouse liver ( Figure 1A). In addition, we discovered that this protein is expressed throughout this organ of mice at postnatal [P] day 2 ( Figure 1A). To further examine how claudin-2 hepatic expression transitions from birth to adulthood, we took advantage of Cldn2-EGFP mice that express a green fluorescence protein (GFP) transgene under the control of Cldn2 regulatory elements . Our immunostaining results showed identical distribution of claudin-2 and GFP in various tissues of wildtype and Cldn2-EGFP mice dissected at P2 or 6 months of age (including the liver [ Figure  . Also, double-immunofluorescence staining of 1-6 months old Cldn2-EGFP livers showed co-expression of GFP with various hepatocyte markers (i.e., HNF4a, Tbx3 and Prox1; Figure 1B), the Zone 3 marker GS ( Figure 1B) and the Zone 2/3 marker Cyp2e1 ( Figure 1B). In contrast, GFP expression was very low or negligible in periportal hepatocytes expressing the cell adhesion protein and Zone 1 marker E-cadherin ( Figure 1B). These results demonstrate that GFP expression faithfully recapitulates the endogenous expression of claudin-2 in many tissues of Cldn2-EGFP mice.
Zones 3 and 2 are reestablished differently in the regenerating liver after acute injury In rodents administered a CCl 4 bolus, the cytochrome P450 enzyme Cyp2e1 metabolizes this compound into highly reactive free radicals that cause cellular damage and necrosis of Zone 2/3 hepatocytes. This process is followed by a regenerative phase encompassing hepatocyte mass recovery, restoration of zonated patterns, and remodeling of the entire organ's architecture (Tanaka and Miyajima, 2016). In spite of the extensive literature on the effects of CCl 4 , our knowledge about the reestablishment of hepatic zonation following acute or chronic exposure to CCl 4 is still limited and most publications in this topic did not examine parenchymal and non-parenchymal cell behaviors simultaneously (Font-Burgada et al., 2015;Pu et al., 2016;Zhao et al., 2019). Furthermore, we do not know if claudin-2 expression is affected following acute CCl 4 exposure or how the expression of this protein is restored in the injury model. To address those issues, we injected Cldn2-EGFP mice with CCl 4 and analyzed their liver 2-7 days post-injection ( Figure 4A). Quantification of ALT/AST serum levels validated the occurrence of liver damage at day two and functional recovery of this organ at day 7 ( Figure 4B) after a CCl 4 bolus.
Our analysis revealed that E-cadherin hepatic expression also changes upon CCl 4 exposure since Zone 2/3 GFP + hepatocytes did not express this protein in control livers ( Figure 4C) and most GFP + hepatocytes expressed E-cadherin 2-5 days after CCl 4 injection ( Figure 4D). Thus, in the CCl 4acutely injured liver E-cadherin expression transiently expands into Zone two during the 'injury phase' and then retracts to its original periportal (Zone 1) domain during the 'repair phase'. Finally, since Tbx3 and claudin-2/GFP proteins colocalize extensively in the uninjured adult liver ( Figures 1C  and 4C) we compared their expression in the Cldn2-EGFP injured liver and uncovered broad overlap of Tbx3 and GFP immunoreactivity throughout the course of recovery (i.e., 2-7 days post-CCl 4 ; Figure 4C). More interesting, we observed that GFPperiportal hepatocytes transiently expressed low-to-moderate levels of Tbx3 2-5 days post-CCl 4 injection ( Figure 4C). This result argues that periportal hepatocytes also respond to local cues during the 'repair phase' that follows CCl 4 -induced injury.
Results of immunostaining showed expression of the GFP lineage-tracer in a few periportal hepatocytes and no GFP expression in Cyp2e1 + hepatocytes (Figure 4-figure supplement 2B) in the CCl 4 -injured Sox9 CreER ;ROSA-EGFP liver. This finding ruled out the contribution of Sox9 + periportal hepatocytes to metabolic Zone 2/3 restoration after CCl 4 -acute damage. Instead, our results indicate that in this injury model metabolic Zone 2 is recovered via proliferation of undamaged hepatocytes and metabolic Zone 3 is induced de novo.
To complement our analysis of Cldn2-GFP livers exposed to CCl 4 , we investigated the composition of the immune infiltrates in tissues harvested 2-7 days after CCl 4 injection ( Figure 5A) using double-immunofluorescence and antibodies recognizing the pan-macrophage marker F4/80, the M1 macrophage marker CD86, and the M2 macrophage marker CD206 (Shapouri-Moghaddam et al., 2018). This analysis showed that most macrophages in Cldn2-GFP control livers were parenchymal Kupffer cells since they expressed F4/80 ( Figure 5B) but not CD86 ( Figure 5C) or CD206 (data not shown). On the other hand, in CCl 4 -injected Cldn2-GFP livers most macrophages infiltrating the central veins represented M1 macrophages that expressed F4/80 and CD86 ( Figure 5B,C) but not CD206 ( Figure 5D and data not shown). Quantitative results demonstrated that the average abundance of perivenous macrophages increased at around day three post-CCl 4 and began to decline afterwards ( Figure 5E). Also, most CD86 + macrophages concentrated around the central veins in CCl 4 -injected livers ( Figure 5C) although in periportal areas of this organ a few isolated CD86 + cells were also found within the hepatic sinusoids ( Figure 5F). This finding argues that perhaps some periportal CD86 + macrophages travel through the sinusoids towards pericentral areas in the CCl 4 -damaged liver.
Endothelial Wnts reestablish claudin-2 zonated expression in hepatocytes following CCl 4 -induced acute injury HECs produce Wnts and other factors that stimulate liver growth (Ding et al., 2010;Hu et al., 2014;Leibing et al., 2018;Preziosi et al., 2018;Zhao et al., 2019) and Wnt ligands that help to repair the architecture of the CCl 4 -injured liver (Zhao et al., 2019). Our finding that most claudin2-GFP + hepatocytes were contiguous to HECs (PECAM-1 + ) during the recovery period that follows a CCl 4 bolus ( Figure 6A) posited that Wnt secretion from HECs help reestablishing metabolic and non-metabolic zonation in this injury model. To test this notion, we administered CCl 4 to Cdh5-CreERT2 (control) and Cdh5-CreERT2;Wls f/f mice at day 0, injected three consecutive daily doses of tamoxifen, and harvested the livers 2-15 days post CCl 4 injection. We uncovered similar increases in circulating ALT/AST levels at day 2 and numerous macrophage infiltrates in hepatic perivenous areas at day 5 in Cdh5-CreERT2 and Cdh5-CreERT2;Wls f/f mice ( Figure 6B). These results validated the occurrence of hepatic damage in both mouse strains. Similar to our previous findings in Cldn2-EGFP mice (Figure 2), Zones 3 (GS + ) and 2 (Cyp2e1 + ) looked almost recovered at day 7 and these areas were completely restored at day 15 in the Cdh5-CreERT2 liver. In contrast, Zone 3 was nearly absent and Zone 2 was significantly reduced in Cdh5-CreERT2;Wls f/f livers at days 5, 7 and 15 post-CCl 4 administration ( Figure 6B). Also, E-cadherin expression retracted to its normal location in periportal Zone 1 in Cdh5-CreERT2 livers 7-15 days post-CCl 4 injection whereas E-cadherin expression was ectopically expanded into perivenous areas in CCl 4 -injured Cdh5-CreERT2;Wls f/f livers dissected at similar time points ( Figure 6B). These results demonstrate that endothelial Wnt secretion is necessary to reestablish Zone 3, to recover the normal size of Zone 2, and to restore E-cadherin expression to Zone 1 hepatocytes, in CCl 4 -acutely injured livers.
HECs are heterogeneous and similar to hepatocytes these cells display zonated distribution of various proteins. For instance, recent studies in adult mouse livers showed that central vein endothelial cells express Wnt2, Wnt9b and Rspo3 transcripts whereas perivenous sinusoidal cells express Wnt2 transcripts (Halpern et al., 2018;Zhao et al., 2019). Also, Kit transcripts (encoding the surface marker CD117) are abundant in HECs located close to the CV and low or absent in HECs located in periportal areas of the adult liver (Halpern et al., 2018). Using immunostaining analysis, we uncovered that the hepatic sinusoids traversing Zones 2 and 3 express CD117 but not those located in periportal liver areas of the adult mouse liver ( Figure 6C). Also, in those tissues the CD117 + sinusoidal cells located in pericentral areas expressed very low levels of Lyve-1 ( Figure 6C). This expression pattern was almost identical in the CCl 4 -injected liver (at day 3) although in the injured organ the LSECs within Zone 2 (Cyp2e1 + ) showed more extensive colocalization of CD117 and Lyve-1 compared to the uninjured liver ( Figure 6C). Since CD117 is expressed more abundantly in percentral/ show that CD117 proteins are restricted to the Lyve-1 + /Lyve-1 LOW sinusoidal endothelium traversing Zones 2/3 in the normal and CCl 4 -injected (day 3) adult liver. (D) Schematic of CD117 + and Lyve1 + HECs isolation. Nonparenchymal liver cells (NPCs) were isolated using a two-step collagenase perfusion method and incubated with CD31-coated Dynabeads. The eluted CD31 + (HEC) fraction was incubated with CD117-coated Dynabeads to isolate CD117 + CD31 + ('pericentral/perivenous') HECs. The unbound fraction from this step was incubated with Lyve-1-coated Dynabeads to separate Lyve + CD31 + hepatic sinusoidal cells from other HECs. (E) QRT-PCR results show comparable Lyve1 transcript expression in CD117 + /CD31 + and Lyve1 + /CD31 + isolates from saline-injected ('control') or CCl 4 -injected (day 3) livers, lower Kit expression in LSECs from injured livers compared to control livers, and higher Wnt2, Wnt9b and Rspo3 expression in LSECs from injured livers compared to control livers. perivenous LSECs than in periportal LSECs, we attempted to isolate these 2 cell populations using CD117-coated magnetic beads and then investigate their Wnt/Rspo expression profile under normal and acute injury conditions. For the 'injury' experiment, LSECs were isolated from mouse livers 3 days after a CCl 4 bolus because at this time point we observed de novo induction of GS expression, expansion of Zone 2 (Cyp2e1 + /claudin-2 + ) and perivenous and periportal hepatocyte proliferation ( Figure 4). Also, for these experiments we included an initial purification step with anti-CD31 antibodies to remove potential contamination of CD117 + immune cells in the injured liver ( Figure 6D). The results of qRT-PCR analysis showed relatively similar Lyve1 expression between CD117 + /CD31 + cells ('pericentral/perivenous LSECs') and Lyve-1 + /CD31 + cells ('periportal LSECs'; Figure 6E). Also, this analysis showed relatively higher Kit expression in CD117 + /CD31 + cells compared to Lyve-1 + / CD31 + cells in both the homeostatic and injured liver. On the other hand, Kit expression was lower in LSECs from CCl 4 -injured livers compared to LSECs from control livers ( Figure 6E) and this result likely indicates that many perivenous LSECs were lost due to extensive tissue damage. Interestingly, the LSECs isolated from CCl 4 -injured livers expressed higher levels of Wnt2, Wnt9b and Rspo3 compared to LSECs from control livers, and this difference was more prominent in the CD117 + /CD31 + LSEC fraction ( Figure 6E). This result was surprising since Rspo3 and Wnt9b transcripts are mainly expressed in pericentral LSECs in the uninjured liver ( Figure 3A and Zhao et al., 2019). Therefore, we postulate that Wnt ligands and Wnt agonists are broadly upregulated in LSECs upon acute liver injury.
Lyve1-cre;Wls f/f mice are refractory to CCl 4 -induced liver damage As we found that the expanding claudin2-GFP + area continuously juxtaposed LSECs (Lyve-1 + ) in Cldn2-GFP livers exposed to CCl 4 ( Figure 7A) and pericentral/perivenous LSECs express Wnt ligands in the CCl 4 -acutely injured liver ( Figure 6E), we investigated whether Wls deletion from LSECs affects zonation restoration in the CCl 4 -injury model. We used Lyve1-Cre;Wls f/f mice for these experiments because we predicted that the remaining Cyp2e1 + hepatocytes ( Figure 7B) should be susceptible to CCl 4 -induced toxicity. CCl 4 administration to Lyve1-Cre (control) mice led to extensive Zone 2 damage, loss of Zone 3 and elevated ALT/AST serum levels 2 days post injection ( Figure 7C), extensive perivenous macrophage infiltration at day 4 ( Figure 7D), and almost complete restoration of tissue architecture and metabolic zonation at day 7 ( Figure 7E). Surprisingly, CCl 4 administration to Lyve1-Cre;Wls f/f mice did not produce any obvious changes in tissue architecture or zonated marker distribution in comparison to Lyve1-Cre;Wls f/f mice injected with saline ( Figure 7B-E; Figure 7-figure supplement 1). Furthermore, Lyve1-Cre;Wls f/f mice injected with CCl 4 had normal ALT/AST serum levels at day 2 ( Figure 7C) and did not show perivenous macrophage infiltrates at day 4 ( Figure 7D). We also injected Cdh5-CreERT2;Wls f/f mice first with tamoxifen (to delete Wls) and 2 weeks later with CCl 4 (to induce Zone 2/3 hepatotoxicity) and then harvested their liver 3 days later. Similar to Lyve1-Cre;Wls f/f mice, in Cdh5-CreERT2;Wls f/f livers administered a CCl 4 bolus we did detect macrophage infiltrates (Figure 7-figure supplement 2B) or observed features indicative of liver damage (Figure 7-figure supplement 2A). Thus, we conclude that the remaining Cyp2e1 + hepatocytes in Lyve1-Cre;Wls f/f livers are protected or refractory to CCl 4 -induced toxicity.

Discussion
The concept of Liver Zonation was proposed almost 40 years ago to explain the spatial separation of metabolic functions in the adult liver (Jungermann and Katz, 1989). Subsequent breakthroughs were the identification of Wnt/b-catenin signaling as a key regulator of metabolic hepatic zonation (Benhamouche et al., 2006;Burke et al., 2009;Burke et al., 2018;Planas-Paz et al., 2016;Sekine et al., 2006;Yang et al., 2014) and the discovery that HECs are a major source of Wnt ligands in the adult liver (Leibing et al., 2018;Planas-Paz et al., 2016;Preziosi et al., 2018;Rocha et al., 2015;Wang et al., 2015). Our study extends those findings by showing that similar regulatory mechanisms control metabolic and non-metabolic zonation in the liver. Specifically, we demonstrate that: 1) endothelial Wnt signaling restricts the expression of a tight junction protein (claudin-2) and the expression of a cell adhesion protein (E-cadherin) to complementary regions in the adult liver; 2) the restoration of metabolic and non-metabolic zonation after CCl 4 -acute injury requires Wnt ligands produced by endothelial cells; and 3) the secretion of Wnt ligands from hepatic sinusoidal cells is dispensable or redundant for the establishment of metabolic and non-metabolic zonation in the newborn liver, and is necessary to preserve zonation patterns in the adult liver.
Our study also introduces a novel transgenic mouse model (Cldn2-EGFP) suitable to investigate changes in claudin-2 expression during development, regeneration and disease. Using those mice,   we discovered a unique distribution of claudin-2 proteins in hepatocytes relative to other membrane proteins and enzymes both at birth and in adulthood. Claudin-2 function is necessary for proper bile flow and bile composition (Matsumoto et al., 2014;Yeh et al., 2010) and the spatiotemporal changes that we uncovered probably reflect how the maturing liver adapts to new metabolic demands and stressors. We also obtained evidence that the mechanism(s) that initiate claudin-2 expression in immature hepatocytes and the mechanism(s) that preserve claudin-2 expression in mature hepatocytes are different. Specifically, we found that Wnt ligand secretion from the hepatic sinusoidal endothelium is necessary to maintain claudin-2 expression in adult hepatocytes and dispensable for the onset of claudin-2 expression in newborn hepatocytes. Our discovery that Wnt/bcatenin signaling regulates claudin-2 expression in mature hepatocytes concurs with a published study showing that b-catenin deletion in adult hepatocytes causes significant downregulation of claudin-2 proteins (Yeh et al., 2010). Investigating how claudin-2 expression is regulated in hepatic cells has clinical relevance because deficiency of this protein causes intrahepatic cholestasis and increases gall stone formation in mice (Matsumoto et al., 2014;Yeh et al., 2010). To further investigate the effects of Wnt signaling on claudin-2 expression, we performed experiments in the hepatocytederived cell line AML-12. Surprisingly, the addition of Wnt2 or Wnt9b alone or in combination with Rspo3 had no demonstrable effect on Cldn2 expression whereas those same factors stimulated the expression of Axin2 (a known Wnt/b-catenin target gene) in AML-12 cells. Possible explanations for those unanticipated results include differences in the cellular context of immortalized AML-12 cells compared to intact hepatocytes, or the requirement of distinct Wnt thresholds and/or specific Wnt ligand combinations to stimulate Axin2 vs. Cldn2 hepatocyte expression.
Another interesting finding in our study is that preserving metabolic and non-metabolic zonation in the adult liver requires Wnt ligand secretion from LSECs. Using the same Lyve1-Cre mouse strain that we used here, other researchers produced Lyve1-Cre;Wls f/f mice and similar to us they described that GS and Cyp2e1 proteins were absent or barely detected in the mutant adult liver (Preziosi et al., 2018). On the other hand, that study did not examine alterations in other zonated proteins in the mutant adult liver or the phenotype of Lyve1-Cre;Wls f/f newborn livers. Furthermore, those authors concluded that Wls was deleted in both CEVs and LSECs in Lyve1-Cre;Wls f/f livers. In contrast, our results of lineage tracing experiments invalidated that cre recombinase was ever expressed in CEVs in our Lyve1-Cre;Wls f/f mice. Whether differences in the genetic background of the mice used in the two studies explain the discrepant results is unclear. On the other hand, our discovery that GS and Cyp2e1 protein expression was relatively normal in the Lyve1-Cre;Wls f/f newborn liver indicates that other cellular sources of Wnts (most probably CEVs) and not LSECs initiate metabolic zonation. Moreover, our results from in situ hybridization experiments and qRT-PCR analysis of isolated LSECs (CD117 + /CD31 + and Lyve-1 + /CD31 + /CD117 LOW ) support a model in which CEVs produce Wnt ligands and agonists (most likely Wnt2, Wnt9b and Rspo3) that induce metabolic Zones 2/ 3 in the perinatal liver, and LSECs secrete Wnt ligands (mainly Wnt2) that preserve metabolic and non-metabolic zonation in the adult liver. Overall, our results support and expand other published data (Ding et al., 2010;Wang et al., 2015;Zhao et al., 2019).
We also took advantage of Cldn2-EGFP mice to analyze simultaneously parenchymal and nonparenchymal cell behaviors in the CCl 4 -acute injury mouse model. Our analysis revealed several interesting new findings in this model, including a small population of intact Cyp2e1 + /GFP + hepatocytes that remained located at the margins of the area of necrosis. Why some Cyp2e1 + hepatocytes survive the injurious insult is not known, but one possibility is that the dose of CCl 4 is too low to access all Zone 2 hepatocytes. On the other hand, some results in our study indicated that the remaining Cyp2e1 + hepatocytes and the surrounding Cyp2e1hepatocytes repaired the damaged Zone 2 through proliferation: 1) those cells expressed the regulator of hepatic progenitor proliferation Tbx3 [(Lüdtke et al., 2009; 2) both Cyp2e1 + hepatocytes and Cyp2e1hepatocytes incorporated EdU in their DNA and retained this label for up to 5 days. This notion is favored by the results of recent studies showing that liver repair mainly occurs via proliferation of perivenous and periportal hepatocytes in different mouse models of liver injury (including the CCl 4 model described here) (Chen et al., 2020;Matsumoto et al., 2020;Sun et al., 2020). Interestingly, we found that in contrast to Cyp2e1 + /GFP + hepatocytes all hepatocytes in pericentral Zone 3 (GS + /GFP + ) were destroyed upon CCl 4 acute exposure and in the next 1-2 days GS expression was induced de novo in a few GFP + hepatocytes located at the front of Zone 2. The number of GS + /GFP + hepatocytes subsequently increased and those cells gradually adjoined the central vein endothelium as the immune infiltrates resolved. These results demonstrate that Zone 3 is specified de novo at the front of an expanding Zone 2 during the repair process that follows CCl 4 -acute liver injury. Similar observations were reported in a recent study in which Axin2 + hepatocytes were genetically ablated and the authors showed complete destruction of Zone 3 and subsequent reappearance of GS + hepatocytes in areas juxtaposed to the CV endothelium (Sun et al., 2020). We hypothesize that Wnt ligands and Wnt agonists produced by intact LSECs are major players in the induction of a new Zone 3 in the CCl 4 -injury model because the new GS + cells were always juxtaposed or near the liver sinusoidal endothelium and, conversely, those cells were physically separated from the central vein endothelium by macrophage infiltrates. Whether a specialized subset of LSECs produces the Wnt ligands and agonists that repair Zones 2 and 3 remains an open question although our results of immunofluorescence and magnetic bead purification experiments suggested that CD117 + /Lyve-1 + LSECs are major sources of Wnt2, Wnt9b and Rspo3 during the repair process that follows CCl 4induced injury.
In summary, our study complements and expands current knowledge on liver zonation and regeneration by showing that multiple Wnt endothelial sources help to orchestrate the spatiotemporal distribution of hepatocyte functions during liver maturation and to promote metabolic zone respecification during liver repair. The advantages we encountered using Cldn2-GFP mice advocate their use in similar investigations of parenchymal and non-parenchymal cell behavior in chronic liver damage or conditions that lead to pathologic claudin-2 expression. Continued on next page

In situ hybridization
Livers were isolated from newborn (P2) or adult mice, cut in slices and fixed in 4% PFA at 4˚C overnight. After embedding in paraffin, 7 mm sections were cut, deparaffinized in xylene, rehydrated in ethanol, and processed for RNA in situ hybridization using the RNAScope 2.5 HD Duplex Kit following the manufacturer's instructions (Advanced Cell Diagnostics

Hepatic endothelial cell (HEC) isolation with magnetic beads
Before HECs isolation, 2 mg of each rat anti-CD31(BD Pharmigen, cat. 550274), rat anti-CD117 (R and D Systems, cat. MAB1356) and rat anti-Lyve-1 (R and D systems, cat. MAB215) antibodies were incubated with 25 ml of magnetic beads (Dynabeads Sheep Anti-Rat IgG, Invitrogen, cat.11035) 1 hr at 4˚C. A Dynamag-15 Magnet (Invitrogen, cat.12301D) was used to isolate the antibody conjugated-beads as per the manufacturer's recommendation. HECs were isolated from mice injected with CCl 4 or saline and sacrificed 3 days later. The liver was perfused through the inferior vena cava first with 20 mL perfusion medium I (1X PBS, 10 mM HEPES, 0.5 g/L KCl, 50 mM glucose, 0.2 mM EDTA, pH 7.4) at 37˚C, followed by 50 mL perfusion medium II (1X PBS, 30 mM HEPES, 0.5 g/L KCl, 50 mM glucose, 1 mM CaCl 2 , 0.5 mg/mL Collagenase H, pH 7.4) at 37˚C. Upon dissecting the liver, the cells were dispersed in cold DMEM medium and passed through a 100 mm strainer. The nonparenchymal cells (NPCs) were separated from hepatocytes by low-speed centrifugation (50 x g, 5 min) at 4˚C. The supernatant containing NPCs was collected and washed twice at 50 x g for 5 min at 4˚C, and pelleted at 170 x g for 10 min. The cell pellets were resuspended in 1 mL red cell lysis buffer (155 mM NH 4 Cl, 12 mM NaHCO 3 , 0.1 mM EDTA) and incubated for 1 min at room temperature (RT), followed by addition of 10 mL isolation buffer (1X PBS Ca 2+ and Mg 2+ free, 0.1% BSA, 2 mM EDTA, pH 7.4). 1 Â 10 7 NPCs were resuspended in 1 mL isolaton buffer and incubated with 25 mL of anti-CD31 Dynabeads for 30 min at 4˚C, After the tube was placed in the magnet for 1 min, the supernatant was discarded, the magnetic beads were washed twice with isolation buffer and the CD31 + cells were incubated with release buffer [Dynabeads FlowComp Mouse CD4, Invitrogen, 11461D] 10 min at RT, eluted and incubated with 25 mL of anti-CD117 Dynabeads for 30 min at 4˚C. After Dynabead binding, the CD117 + /CD31 + fraction was isolated as indicated above and lysed in TRIzol (1 mL) for RNA isolation. The unbound supernatant was incubated with anti-Lyve-1 Dynabeads and the Lyve1 + /CD31 + /CD117 low cells were isolated as previously indicated.

Statistics and quantitative analysis
The data represent the mean ± SEM and P value were calculated by two-tailed unpaired Student's ttest ( Figures 2E, 3C, D, 4F and 6E) or one-way ANOVA ( Figures 1D, 4E, 5E, 6B and 7C) or twoway ANOVA ( Figures 3E, F and 4B) by the GraphPad Prism 8.0 software. NS, not significant (p>0.05), *p<0.05, **p<0.01, ***p<0.001. Each quantitative experiment was repeated at least three times. We considered biological replicates as those animals or tissues subjected to the same experimental test, and technical replicates as individual samples or tissues subjected to the same analysis.

Study approval
All animal experiments were performed in accordance with protocols reviewed and approved by the Institutional Animal Care and Use Committee at Northwestern University. The animal welfare assurance number for this study is A3283-01.
experiments, and the GENSAT BAC Transgenic project (Rockefeller University, NY) for the Cldn2-GFP mice. CONACYT awarded a fellowship to Dr. Martinez-Ramirez The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.