Mice depleted for Exchange Proteins Directly Activated by cAMP (Epac) exhibit irregular liver regeneration in response to partial hepatectomy

The exchange proteins directly activated by cAMP 1 and 2 (Epac1 and Epac2) are expressed in a cell specific manner in the liver, but their biological functions in this tissue are poorly understood. The current study was undertaken to begin to determine the potential roles of Epac1 and Epac2 in liver physiology and disease. Male C57BL/6J mice in which expression of Epac1 and/or Epac2 are deleted, were subjected to partial hepatectomy and the regenerating liver was analyzed with regard to lipid accumulation, cell replication and protein expression. In response to partial hepatectomy, deletion of Epac1 and/or Epac2 led to increased hepatocyte proliferation 36 h post surgery, and the transient steatosis observed in wild type mice was virtually absent in mice lacking both Epac1 and Epac2. The expression of the protein cytochrome P4504a14, which is implicated in hepatic steatosis and fibrosis, was substantially reduced upon deletion of Epac1/2, while a number of factors involved in lipid metabolism were significantly decreased. Moreover, the number of Küpffer cells was affected, and Epac2 expression was increased in the liver of wild type mice in response to partial hepatectomy, further supporting a role for these proteins in liver function. This study establishes hepatic phenotypic abnormalities in mice deleted for Epac1/2 for the first time, and introduces Epac1/2 as regulators of hepatocyte proliferation and lipid accumulation in the regenerative process.

both at mRNA and protein level, that the mouse liver only expresses the Epac2C isoform 8 . Epac1 transcripts were identified in sinusoidal endothelial cells, Küpffer (KC) cells and HSC, whereas Epac2C was found solely in hepatocytes (Fig. 1b). Thus, none of the cell types analyzed were found to co-express Epac1 and Epac2.
Deletion of Epac1/2 accelerates DNA replication after PH. To gain insight into the functional roles of Epac1/2 in the liver, wt, Epac1 −/− , Epac2 −/− and Epac1/2 −/− mice were subjected to PH and euthanized at four post-operative time points (26 h, 36 h, 74 h and 168 h). Mice depleted for Epac1/2 displayed no adverse phenotypes in response to PH compared to wt mice (i.e. no change in post-operative hepatocellular necrosis or mortality), and the resected liver-to-body weight ratio was similar in all genotypes ( Supplementary Fig. S3a), as was the restitution of hepatic index ( Supplementary Fig. S3b). qPCR was performed to determine whether PH affects the expression of Epac1/2 in wt mice. As shown in Fig. 2a, Epac2c mRNA expression in total liver was increased at 26 h and 36 h post-PH, but then returned to basal levels. This finding correlates with increased Epac2C protein expression as determined by proteomic analyses (Supplementary file 2). In contrast, Epac1 mRNA levels decreased at 26 h post-PH (Fig. 2b). However, confounding the relevance of this finding is the very low level of Epac1 mRNA in total liver (see legend to Fig. 2 for ct-values), as also evident from undetectable levels of Epac1 in proteomic analyses of total liver (Supplementary file 2).
To determine whether the proliferative capacity after PH was affected as a consequence of deleting Epac1/2, BrdU incorporation in hepatocytes was analyzed to quantify DNA synthesis. In wt mice, BrdU-positive hepatocytes were detectable at 26 h, peaked at 36 h and returned to basal levels 74 h after PH, similar to previously reported 28 . Interestingly, the number of BrdU-positive hepatocytes was substantially increased in Epac1 −/− , Epac2 −/− and Epac1/2 −/− mice compared to wt mice 36 h post-PH (Fig. 3a), demonstrating that more hepatocytes had entered the S-phase. The increased BrdU incorporation was accompanied by an increased number of DAPI stained nuclei per area 36 h post-PH (Fig. 3b). Since the hepatocyte size was similar in wt and Epac1/2 −/− mice ( Supplementary Fig. S3c), these results indicate increased hepatocyte proliferation in Epac1 −/− , Epac2 −/− and Epac1/2 −/− mice, and not increased hepatocyte hypertrophy, which is often observed after PH 29 . The increased proliferation of hepatocytes in Epac1/2 −/− mice was a result of PH, as DAPI staining revealed similar numbers of hepatocytes in resected tissue sections of all genotypes before PH, and because BrdU incorporation was equal in response to sham operation in all genotypes ( Supplementary Fig. S3d).
Proteomic analyses indicated no changes in the expression of cyclin dependent kinases (Cdks) in Epac1/2 −/− mice compared to wt mice (Supplementary dataset file 2). We were unable to detect other major cell cycle factors The CT values for resected tissue (4.39 ± 0.12 for Epac2C and 12.29 ± 0.38 for Epac1) were set to 1. Data are expressed as mean ± SD of three separate experiments performed in triplicates. One-way ANOVA with Dunnett's adjustment for multiple comparisons was used to determine statistical differences. ***p < 0.001 and ****p < 0.0001: mRNA levels in wt livers pre-PH compared to mRNA levels in wt livers at the different postoperative time points ( The regenerating liver of Epac1/2 −/− mice display early loss of fat vacuoles. Mice accumulate hepatic lipid droplets in response to PH, and this condition of steatosis is particularly evident 20-40 h post-PH 32 . Oil Red O (ORO) staining of the right and caudate lobes that remained after PH revealed extensive lipid accumulation in wt mice at 26 h and 36 h post-PH, followed by a return to basal levels at later time points (Fig. 4). Lipid droplets were also observed in Epac1 −/− , Epac2 −/− and Epac1/2 −/− mice 26 h post-PH. The size, shape and distribution of the microvesicular fat vacuoles were comparable in all genotypes at this time point, although quantitative determination of the ORO staining revealed notable individual variations between animals within the same experimental group (Fig. 4b). This was especially evident in Epac1/2 −/− mice where some individuals exhibited moderate steatosis 26 h post-PH, while others presented with a phenotype completely devoid of fat vacuoles (Fig. 4b). Of note, all Epac1/2 −/− mice presented with a striking absence of fat vacuoles 36 h post-PH (Fig. 4a). Reduced fat accumulation was also evident in Epac2 −/− mice at both time points (26 h and 36 h post-PH) compared to wt mice (Fig. 4).
Altered lipid profiles and protein expression in the absence of Epac1/2. Plasma lipid profiling revealed decreased basal (pre-PH) concentrations of circulating total cholesterol in Epac1 −/− , Epac2 −/− and Epac1/2 −/− mice compared to wt mice (Table 1), as previously described in Epac1 −/− mice 33 . The reduction in total cholesterol correlated with reduced levels of high-density lipoprotein (HDL) cholesterol, as neither low-density lipoprotein (LDL) cholesterol nor unesterified cholesterol levels were changed (Table 1). In accordance with the reduced level of HDL cholesterol, circulating phospholipid was also reduced in all Epac1/2 knockout models prior to PH. Moreover, the basal levels of triglycerides (TG) was elevated in plasma of Epac1/2 −/− mice (Table 1), but  The median and left lateral lobes resected during surgery were used as control tissue in these experiments. Data are shown as box and whiskers with median and min to max whiskers, n = 4-8 mice/group post-PH and n = 17-23 mice/group pre-PH. Two-way ANOVA with Bonferroni's adjustment for multiple comparisons was used to determine statistical differences between genotypes within each time point. *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001. F-statistics: F(12,149) = 5.152, p < 0.0001. in LDL cholesterol was not significant (Table 1). Of note, Epac1/2 −/− mice had a significantly higher level of plasma LDL cholesterol than wt and Epac2 −/− mice post-PH, indicating increased LDL cholesterol secretion or decreased LDL cholesterol uptake in peripheral tissues. The concentration of circulating free fatty acids (FFA) was similar before and 36 h after PH in all genotypes (Table 1), demonstrating that the concentration of circulating FFA that peaks around 20 h post-PH in mice 34 , returns to normal also in the absence of Epac1/2.
The hepatic lipid profiles were nearly identical in the different genotypes in mice not subjected to PH ( Table 1). The only variation found was that Epac1 −/− mice had lower levels of phospholipids compared to Epac1/2 −/− . However, taken together, these results indicate that under resting conditions, lipid metabolism is adequately maintained in the absence of Epac1/2. In response to PH, increased hepatic levels of TGs and cholesterol were observed in wt mice as reported 35 . Similarly, cholesterol was significantly increased in Epac1 −/− ad Epac2 −/− mice ( Table 1). TG levels were also elevated in these genotypes, but due to large variations between individuals within the same genotype (Epac1 −/− and Epac2 −/− ), the differences where not statistically significant. As also evident from the ORO staining (Fig. 4), Epac1/2 −/− mice exhibited lower levels of TGs in the livers post-PH, but the large individual variations within each genotype resulted in non-significant statistical results. In contrast to the other genotypes, hepatic cholesterol was not increased in Epac1/2 −/− mice after PH, also contributing to an overall reduced mass of neutral lipids in this group.
The dramatic effect on lipid accumulation in Epac1/2 −/− mice 36 h after PH, but milder effects in mice deleted for either Epac1 or Epac2 ( Fig. 4 Table 1. Altered lipid profile in Epac −/− mice. Enzymatic calorimetric lipid profiling of plasma and liver tissue isolated from wild type (wt), Epac1 −/− , Epac2 −/− and Epac1/2 −/− control mice and PH operated mice 36 h post surgery. Data are presented as mean ± SD, n = 3-10 mice/group. Two-way ANOVA with Bonferroni's adjustment for multiple comparisons or the non-parametric Kruskal-Wallis test with Dunn's multiple comparison test was used to determine differences between genotypes and treatment groups for each factor. www.nature.com/scientificreports www.nature.com/scientificreports/ in Epac1/2 −/− mice compared to wt mice (54 increased and 59 decreased). Gene ontology (GO) enrichment analyses of the proteins with increased expression pre-PH in Epac1/2 −/− mice relative to wt mice indicated over-representation of proteins connected to sulfation and sulfotransferase activity, and biosynthetic processes related to, among other molecules, cholesterol biosynthesis and steroid metabolism ( Fig. 5a; see Supplementary  Table 1a for an overview of the proteins identified for each GO-term). The proteins with reduced expression in Epac1/2 −/− mice compared to wt mice pre-PH were mainly associated with lipid metabolism, peroxisome biology and oxidation-reduction processes (Fig. 5b, and Supplementary Table 1b). Post-PH, again 113 proteins were differentially expressed in Epac1/2 −/− mice compared to wt mice (38 increased and 75 decreased). GO enrichment analyses of the proteins with decreased expression post-PH predicted enrichment of similar or related pathways and processes as before PH, but interestingly, the statistical prediction was considerably strengthened for the terms related to peroxisome biology ( Fig. 5c; see Supplementary Table 2 for an overview of the proteins identified for each GO-term). There were no significantly enriched GO-terms for the proteins that were increased in Epac1/2 −/− mice compared to wt mice post-PH. In accordance with previous studies 36 , PH altered the hepatic protein composition in wt mice, resulting in enrichment of GO terms related to metabolic processes, mainly lipid metabolism, the inflammatory response and regeneration ( Supplementary Fig. S6). The expression of proteins linked to lipid metabolism was also changed in Epac1/2 −/− post-PH, and interestingly, the GO analyses predicted enrichment for the terms "Negative regulation of gluconeogenesis" and "Very-low density lipoprotein particle" in the absence of Epac1/2 ( Supplementary Fig. S6).
Of particular interest is that the expression of cytochrome P4504A14 hydroxylase (Cyp4a14) was severely affected in Epac1/2 −/− mice. This enzyme catalyzes omega-hydroxylation of medium-chain fatty acids, and is implicated in hepatic steatosis and fibrosis 37 . The proteomic analyses demonstrated a 10 and 14 fold repression of Cyp4a14 in Epac1/2 −/− mice compared to wt mice before and after PH, respectively (Fig. 6a,b). IHC and immunoblotting analyses verified the decreased expression of Cyp4A14 in mice deleted for Epac1, Epac2 or Epac1/2 ( . The antibody used in IHC and immunoblotting recognizes Cyp4a10 and Cyp4a12, in addition to Cyp4a14. Thus, the IHC and immunoblotting results do not perfectly mimic the reduction in Cyp4A14 expression observed in the proteomic analyses. Cyp4a12 was downregulated 1.7 fold in Epac1/2 −/− mice both before and after hepatectomy (Fig. 6), and Cyp4a10 was downregulated 2 and 4 fold in Epac1/2 −/− mice pre-and post-PH, respectively (Fig. 6). The IHC analyses demonstrated that in wt mice, the highest expression of Cyp4A14 was in the centrilobular region 38 . Epac1 −/− , Epac2 −/− and Epac1/2 −/− mice exhibited a distribution pattern with fewer centrilobular Cyp4A14 positive cells, a weaker staining intensity and a localization that included Cyp4A14 positive cells in the transition-and periportal regions pre-PH (Fig. 7).
Cyp4a14 is a target gene for peroxisome proliferator-activated receptor alpha (PPARα) 39 , and it is notable that a considerable number of the proteins that were differentially expressed in Epac1/2 −/− mice compared to wt mice (both pre-and post-PH) are known PPARα target genes. In particular, enzymes involved in fatty acid degradation were generally decreased in Epac1/2 −/− mice as opposed to wt mice (Table 2). However, immunoblotting analyses demonstrated no effect on PPARα expression as a consequence of Epac1/2 deletion (Supplementary Fig. S9; see Supplementary Fig. S10 for full-length blots) and the proteomic analysis did not provide insight into a potential effect of PPARα expression in Epac1/2 −/− mice, as we were unable to detect PPARα in this analysis. Cytochrome P4507α hydroxylase (Cyp7α1), was the protein found to be most significantly increased in Epac1/2 −/− mice post-PH compared to before surgery (5-fold) (Fig. 6b). This was in sharp contrast to wt mice where the expression of CYP7α1 was decreased in response to PH, in accordance with the literature 40 (Supplementary dataset file 2). Cyp7α1 is also a PPARα target gene 39,41 , and catalyzes the first reaction in the bile acid biosynthetic pathway 42 . The expression of other enzymes involved in this process (i.e. Akr1d1: Aldo-keto reductase family 1 member D1 and HSD17b4: Hydroxysteroid 17-beta dehydrogenase 4) was also altered in Epac1/2 −/− mice. Of note is that the expression levels of the different PKA subunits (i.e. Prkar1a Prkar2a, Prkaca and Prkacb) were not changed in Epac1/2 −/− mice (Supplementary dataset file 2), and neither were the expression of the different isoforms of the Epac1/2 downstream targets Ras-related protein Rap1/2 (Rap1a, Rap1b, Rap2a:Rap2c;Rap2b). Recently, several studied have shown that Epac1/2 is localized to the mitochondria [43][44][45][46] . In this regard, it is interesting to note that a substantial number of the proteins with decreased expression in Epac1/2 −/− mice that are associated with lipid metabolism and oxidation-reduction processes, are mitochondrial proteins (Supplementary Table S3).
The number of KCs is affected in Epac1 −/− and Epac2 −/− mice. It is intriguing that deletion of either Epac1 or Epac2 alone causes phenotypes in response to PH, and that some of the effects are reinforced when both factors are deleted. These findings suggest that also Epac1 expressing cells, and not only hepatocytes, are affected in the livers of the knockout models. To gain further insights into the potential effects on nonparenchymal cells, we quantified the relative numbers of KCs and HSCs by IHC. Both cell types are known to exert stimulatory and inhibitory effects on hepatocyte proliferation following PH 47,48 . We observed no significant differences between genotypes with regard to the number of platelet derived growth factor beta (PDGFβ) -positive HSCs (Fig. 8a). However, the number of F4/80-positive KCs was increased in Epac1 −/− mice pre-PH, and decreased in Epac2 −/− mice 36 h post-PH (Fig. 8b).

Discussion
The mammalian response to PH has been extensively studied, but the details of how the signaling cascades that control liver regeneration integrate, remain incompletely understood. It is known that cAMP and PKA are involved, and the production of cAMP is tightly controlled during the process [49][50][51][52] . Moreover, a delay in regeneration is observed in mice deleted for the cAMP-regulated transcription factor CREM (cAMP responsive element modulator) 3 . In the present study, we demonstrate for the first time that the cAMP binding proteins Epac1/2 are involved in regulating liver regeneration in an in vivo model. Deletion of Epac1 or Epac2, or both factors together, www.nature.com/scientificreports www.nature.com/scientificreports/ caused phenotypes manifested in increased lipid turnover and DNA synthesis early in the regenerative process. Moreover, we found that expression of Epac2C was upregulated in the regenerating liver, that the number of KCs cells was altered and that the liver proteome of Epac1/2 −/− mice differed from wt mice, a disparity that was www.nature.com/scientificreports www.nature.com/scientificreports/ substantiated in response to PH. Taken together, it appears that loss of Epac1/2 is compatible with normal liver functions in mice living in a protected environment, but that these factors become important in the regulation of cellular responses in stressful situations. We used global knockout models in this study, and cannot disregard possible indirect effects of Epac1/2 deletion on liver regeneration, and thus, future analyses to decipher the molecular Figure 6. Differentially expressed proteins in Epac1/2 −/− mice compared to wt. The proteins that were differently expressed in Epac1/2 −/− mice compared to wt mice (a) pre-and (b) post-PH are listed. Blue bars indicate proteins that were decreased in Epac1/2 −/− mice compared to wt mice, and red bars indicate proteins that were increased in Epac1/2 −/− mice compared to wt mice. Statistics: two-tailed student's two-sample t-test and Z-statistics for FC, only when p < 0.05 for both it was considered significant. www.nature.com/scientificreports www.nature.com/scientificreports/ roles of Epac1 and Epac2 in different hepatic cell lines should include cell-specific knockout models. An example of a possible indirect effect is via blood platelets. It is clear that platelets and platelet-derived factors have central roles in liver regeneration after PH [53][54][55] , and in this regard it is interesting to note that we previously demonstrated that Epac1 −/− mice (but not Epac2 −/− mice) display a bleeding phenotype that is partly caused by altered platelet maturation and function 56 . The fact that some of the observed phenotypic traits were substantiated in Epac1/2 −/− mice compared to single knockouts, indicates that these proteins have distinct, but convergent roles in the regenerative process, and that lack of one factor (and both) causes defective cell communication and integration of signals. This notion is supported by the cell specific and mutual exclusive expression of Epac1 and Epac2 in the liver, and by the finding that the number of Epac1 expressing KCs is altered not only in Epac1 −/− mice, but also in Epac2 −/− mice post-PH. Epac1 exerts anti-inflammatory effects in different biological systems 57 , and at present, it is unclear whether the increased number of F4/80-positive KCs observed in Epac1 −/− mice results from increased recruitment of circulating macrophages due to an increased inflammatory state in the absence of Epac1, or caused by intrinsic effects of the liver. Interestingly, the pronounced increase in TNFα that we observe in Epac1 −/− mice before PH is in agreement with an enhanced recruitment of circulating monocyte-derived macrophages, which is known to secrete pro-inflammatory cytokines, whereas the pro-regenerative liver resident KCs secrete anti-inflammatory cytokines 58 .   www.nature.com/scientificreports www.nature.com/scientificreports/ The discovery that the number of KCs is changed in Epac2 −/− mice 36 h post-PH might imply that the interaction between hepatocytes and KCs is disturbed. We have limited knowledge of how hepatocyte mediated signaling affects KC replenishment post-PH. However, it has been suggested that dysregulated secretion of granulocyte-macrophage colony-stimulating factor (GM-CSF) from hepatocytes interferes with, or delays KC replenishment post-PH 59,60 . This is interesting considering the suggested role for GM-CSF in hepatic lipid homeostasis 61 , and growing evidence demonstrating that KC renewal is affected by the metabolic state of the liver 62 . It is yet to be determined whether GM-CSF mediated signaling is affected in the absence of Epac2C.
Rap1 GTPase, the main downstream Epac1/2 effector, plays an important role in regulating cell proliferation. However, the roles of Epac1/2 in cell proliferation have remained elusive, as both pro-mitotic and anti-mitotic effects have been ascribed to these proteins depending on cellular context 63 . The increased BrdU incorporation observed in Epac1 −/− , Epac2 −/− and Epac1/2 −/− mice at 36 h post-PH in the present study, indicates that the absence of Epac1/2 causes an accelerated first round of hepatocyte proliferation. In mice, coordinated waves of hepatocyte proliferation occur in response to PH, with the first round being the most robust 64 . The expression of cell cycle regulators are expressed within hours after PH, and interestingly, it has been demonstrated that the relative expression of the different cell cycle components is rather consistent throughout the liver regrowth, despite fluctuations in DNA synthesis during the process 65 . In line with this, although we detected increased proliferation at 36 h post-PH, we did not observe increased expression of Cyclin D1, Cyclin E1or cyclin dependent kinases (Cdks) in mice deleted for Epac1/2. However, we can not rule out that the levels of these factors differed between wt and Epac1/2 knockout models at time points earlier than 26 h and therefore escaped detection in our experimental set-up. We found that the expression of TNFα is lower in Epac1/2 −/− mice, and partly in Epac2 −/− mice compared to wt mice at the peak of DNA-synthesis (36 h post-PH). TNFα triggers the priming phase in hepatocyte proliferation after PH, whereas this factor is suppressed at later stages 66 . Thus, the observed decreased expression of TNFα in Epac1/2 −/− mice substantiates the idea that hepatocyte proliferation occurs more rapidly in mice lacking Epac1/2. Taken together, we suggest that the increased number of BrdU-positive hepatocytes in Epac1 −/− , Epac2 −/− and Epac1/2 −/− mice 36 h post-PH, demonstrate that the Epac proteins restrain hepatic cell proliferation under normal conditions.
A hallmark of the early stages of liver regeneration in response to PH is the accumulation of lipids, which contributes a greater portion of the energy substrates necessary to complete regrowth 67 . The finding that Epac1/2 −/− mice are completely devoid of lipid droplets 36 h post-PH indicates regulatory roles for Epac1/2 in this process in wt mice. Despite large individual differences, the ORO-staining at 26 h points to the fact that Epac1/2 −/− mice were capable of accumulating lipids at early time points after PH. We propose that ablation of Epac1/2 causes a shift in the regeneration process with a more rapid lipid turnover accompanied with increased hepatocyte proliferation. These irregularities did affect neither the overall regeneration process nor the survival of the animal, indicating that the role of Epac1/2 is to balance the regulatory mechanisms that control the process. This is consistent with several publications demonstrating that normal lipid accumulation in response to PH is not a prerequisite for normal regeneration 34,68 .
The proteomic analyses presented in this study disclosed that the expression of Cyp4A14 is dramatically decreased in Epac1/2 −/− mice, both pre-and post-PH. This enzyme, and the human homolog (CYP4A hydroxylase), catalyze omega-hydroxylation of medium chain fatty acids and arachidonic acids 69 . Interestingly, Cyp4A14 expression is increased in murine models of spontaneous nonalcoholic fatty liver disease (NAFLD) [70][71][72] , as well as in livers of patients with NAFLD 37 . Moreover, overexpression of, or genetic ablation of Cyp4A14 in mice correlates with increased and reduced hepatic fat accumulation, respectively, in response to a high fat diet 37 . The result that Cyp4A14 is strongly repressed in livers from Epac1/2 −/− mice both before and after hepatectomy provides a link between Epac1/2, Cyp4a14 and the regulation of hepatic lipid turnover. NAFLD is displayed as a range of diseases that can develop to non-alcoholic steatohepatitis and further to fibrosis 73 . Cyp4a14 also plays a role in the development of fibrosis, as diet induced portal and perisinusoidal fibrosis is attenuated in mice depleted for Cyp4A14 expression, as is the expression of fibrosis-related genes in HSC (activation of HSC is central in the development of hepatic fibrosis) 37 . In this context it is of interest that Epac1 has been proposed to exhibit anti-fibrotic functions, and that several studies point to Epac1 as an integrator of pro-and anti-fibrotic signals in HSC 13,17,18 . Whereas the pro-fibrogenic factor TGFβ1 decreases Epac1 expression in HSC, overexpression of Epac1 inhibits TGFβ1-induced production of collagen 13 . Expression of Epac1 is also repressed in fibrotic livers of mouse and human, and stimulation of Epac1 expression in HSC partly restores carbon tetrachloride induced fibrosis in mice 17 . The effects of high fat feeding in mice deleted for Epac1/2 is not yet clear, as conflicting results are obtained from different Epac knockout models 23,33,74 . A recent study describes increased obesity and fat accumulation in adipose tissue and liver in male mice deleted for Epac2A 74 . However, since the model used in this study 74 expresses normal levels of Epac2C in hepatocytes, it does not provide information about the intrinsic roles for Epac2C in the liver. On the other hand, it does imply that Epac2A, acting outside of the liver, affects hepatic lipid accumulation indirectly, prompting future studies on hepatocyte specific Epac2C knockouts 69 .
Several mouse models that lack expression of Epac1/2 have been developed 23,33,[75][76][77][78] . A common feature of these models is that deletion of either Epac1 or Epac2, or both, does not produce gross anatomical or physiological abnormalities, but that exposure to stress provokes phenotypes in a diversity of physiological processes, as we describe herein for the liver 23,33,56,75,76,[78][79][80][81][82][83] . Collectively, these studies suggest that Epac1/2 may have evolved to balance cellular responses in challenging situations, and implicate Epac1/2 in a number of diseases. Epac1/2 are considered promising drug targets, and Epac2 has been suggested to be the target for both incretin-related drugs and sulfonylureas, which are widely used to treat type 2 diabetes 84 . Currently, considerable effort is invested in developing compounds that specifically act on either Epac1 or Epac2 to tailor treatment and reduce side effects 26,85 . The roles of Epac1/2 in liver physiology disclosed in the present study hold promise for that compounds acting via Epac1/2 may be used in the future to treat liver diseases. (2019) 9:13789 | https://doi.org/10.1038/s41598-019-50219-8 www.nature.com/scientificreports www.nature.com/scientificreports/

Materials and Methods
Animals. Male mice (C57BL/6J background, 8-12 weeks) were used in this study. Epac1 −/− and Epac2 −/− mice were generated using the Cre-LoxP targeting strategy described elsewhere 56,[81][82][83] . Of note, the targeting strategy used ablates the expression of all Epac2 isoforms (i.e. Epac2A, Epc2B and Epac2C). Mice deleted for Epac1 and Epac2 (Epac1/2 −/− ) were generated by crossing of Epac1 −/− and Epac2 −/− mice. Wt mice used in the study (originally from Taconic, Denmark) were littermates of Epac1 −/− and Epac2 −/− mice or, mice bred in the same room. The origin of the C57BL/6JBomTac mice is described at https://www.taconic.com/mouse-model/b6jbom. The mice were housed in groups of three to five and bred at the animal facility at Haukeland University Hospital and maintained under standard conditions with a temperature of 21 °C ± 0.5 °C, humidity of 55% ± 5%, lighting of 150 lux and 12 hours light/dark cycle (7 am/7 pm). Animals were housed in IVC II cages with food (RMI (E)) and water ab libtium. All animal protocols were approved by the Norwegian Animal Research Authority and performed according to the European Convention for the Protection of Vertebrates Used for Scientific Purposes. The Animal Care and Use Programs at University of Bergen are accredited by AAALAC international.
Hepatectomy. 2/3 partial hepatectomy (PH) or sham laparotomy was performed as described 86 . At the selected post-operative endpoints (26 h, 36 h, 74 h and 168 h), the mice were euthanized by CO 2 -suffocation, with administration of BrdU (50 mg/kg IP) 2 h before sacrifice. Tissue removed at time of surgery (left lateral and median lobes) and regenerated tissue (right lateral and caudate lobes) removed at selected endpoints were flash-frozen on liquid nitrogen, or fixed in 4% paraformaldehyde (PFA; 10x w/v) for 18-24 h, embedded in paraffin and sectioned (4 μm). Trunk blood was obtained by cardiac puncture, collected into prechilled EDTA-tubes and blood plasma prepared. See Supplementary S1 Materials & Methods, for details on H&E, ORO staining, hepatocyte size quantification, lipid profiling, immunohistochemistry and immunoblotting. slipped. DAPI-stained sections were not blocked, but directly counterstained and mounted. The stained sections were visualized with Zeiss Axioplan2 microscope and AxioVision version 4.5 Program (Zeiss), hepatocyte proliferation was determined by manually counting positive BrdU nuclear staining. For each mouse, ten 40x microscopic fields (in an area of 35.2 mm 2 ) in three different sections were counted blindfolded by two different researchers. Likewise, DAPI staining was counted manually in ten 40x microscopic fields in one section from each mouse.

Real time (RT) and quantitative (q) PCR. Primary cell isolation and fluorescence-activated cell sort-
ing. Hepatocytes and HSC were isolated as described previously 8,87 . Liver resident macrophages (KCs) were extracted as described previously 88 . In brief, mouse liver was perfused through the inferior vena cava with PBS. Livers were excised, minced and digested in solution I (0.8 mg/ml collagenase V, 0.625 mg/ml collagenase D, 1 mg/ml Dispase and 30 μg/ml DNase). The suspension was shaken on an orbital shaker at 37 °C for 25 min, with additional vigorous manual shaking every 5 min. Digest was strained through a 100 μm cell strainer. Following 2 × 5 min centrifugation at 300 g RBC lysis buffer was added for 3 min to lyse red blood cells. Cells were then centrifuged at 300 g for 5 min, resuspended in FACS buffer (2% FBS (Gibco) and 2 mM EDTA in Dulbecco's phosphate-buffered saline no calcium, no magnesium (Gibco)) and filtered through a 35 μm strainer. Cells were blocked with 10% mouse serum (Sigma) containing 1% Purified anti-mouse CD16/32 (Biolegend) for 10 min. Cells were then incubated with fluorescent conjugated antibodies (FITC anti-mouse CD45.2, PE/Cy7 anti-mouse F4/80, APC/Cy7 anti-mouse/human CD11b, PE anti-mouse CD3, PE anti-mouse CD19, PE anti-mouse Ly-6G, APC anti-mouse CD31 and PE Rat Anti-Mouse Siglec-F) or isotype control at 4 °C for 30 min in the dark. Cells were washed, resuspended in FACS buffer and sorted using a FACS Aria II (BD Biosciences, USA). Following a positive CD45.2 gate and dump of CD3, CD19, Ly-6G, Siglec-F, and CD31 positive KCs were defined as the F4/80 high, CD11b intermediate population.
To isolate liver sinusoidal endothelial cells (LSEC), mouse liver was perfused through the inferior vena cava with PBS. Livers were excised, minced and digested in a solution containing 5 mg/ml collagenase I and 11μg/ml DNase. This cell suspension was shaken at 37 °C for 20 min. Digest was strained through a 70 μm cell strainer. Following 7 min centrifugation at 400 g, RBC Lysis buffer was added for 5 min. Cells were then centrifuged at 400 g for 7 min, resuspended in FACS buffer and filtered through a 35 μm strainer. Cells were blocked with 10% mouse serum and 1% CD16/32 for 10 min. Cells were then incubated with fluorescent conjugated antibodies (PE/ Cy7 anti-mouse CD45 and APC anti-mouse CD31 clone. 390) or isotype control at 4 °C for 30 min in the dark. Cells were washed, resuspended in FACS buffer and sorted using a FACS Aria II. Endothelial cells were sorted from the CD31 positive, CD45 negative population.
RNA was extracted from the isolated cells using the RNeasy Micro kit (74004, Qiagen, Germany), and converted to cDNA using the SuperScript ® III First-Strand Synthesis SuperMix for qRT-PCR kit (Invitrogen, Waltham, MA, USA). The RT-PCR reaction was carried out as described in 8 . For qPCR analyses, RNA was prepared from total liver and converted to cDNA using the GenElute Mammalian Total RNA miniprep Kit, and