**4. Discussion**

Little is known about how liver fibrosis influences hepatic zonation. Here, we show that fibrosis of different etiologies causes 'periportalization', which means that the entire liver lobule including the pericentral region adopts a periportal pattern of gene expression (graphical abstract). We first observed this phenomenon in the mouse model of CCl4 established here, where severe fibrosis was induced by repeated induction of pericentral liver necrosis over a period of one year. Recently, several genome-wide studies have identified genes that show a preferentially pericentral or periportal expression pattern [31–33]. We established a consensus list of periportal and pericentral genes that overlapped in at least two of the three studies. The genome-wide analysis showed that pericentral genes are enriched among the genes downregulated by CCl4, while the periportal genes are enriched among the upregulated genes. This phenomenon occurred already at the earliest analyzed time point of two months.

Analysis of selected pericentral genes by immunostaining and qRT-PCR showed that the downregulation of pericentral genes occurred in two phases. Phase 1 until month 6 is characterized by 'central-to-central bridging'. In this period, the pericentral area with positive immunostaining of e.g., cytochrome P450 enzymes becomes narrower but areas of contact between neighboring pericentral regions are maintained. In this first phase, RNA levels of the analyzed pericentral genes show a statistically significant but moderate, less than 2-fold decrease. Phase 2 occurs between months six and twelve and is characterized by a massive formation of fibrotic streets. In this period, immunostaining of cytochrome P450s and glutamine synthetase is almost completely lost and RNA levels decrease by more than 10-fold.

Immunostaining of the periportal genes arginase 1 and CPS1 shows that they begin to be newly expressed in the pericentral region during CCl4 treatment. In controls, they are expressed in periportal and midzonal hepatocytes but not in a narrow region around the central vein. While pericentral proteins retreat, arginase 1 and CPS1 advance towards the central vein until finally all hepatocytes express these periportal genes. The extension of periportal enzymes into the pericentral region is accompanied by a moderate but statistically significant increase in RNA levels (e.g., of CPS1) at months 2 and 6. However, between months 6 and 12, expression of several periportal genes decreases again. This is most likely explained by the massive formation of fibrotic streets that contain myofibroblasts, macrophages and further inflammation-associated cells that do not express (periportal) liver genes. Therefore, periportal liver genes will be diluted by the RNA of the cells forming fibrotic streets. Nevertheless, immunostaining suggests that expression levels, e.g., of arginase 1 and CPS1, do not

decrease in the remaining hepatocytes of the regenerative nodules at month 12. Rather, the hepatocyte fraction of total liver mass decreases.

The loss of pericentrally expressed metabolizing enzymes has different consequences. One of them is an increase of ammonia concentrations in the liver vein blood and the systemic circulation. During the passage through the fibrotic, 1-year CCl4-treated liver, ammonia concentrations still decrease from ~600 μg/dL concentrations in the portal vein to ~180 μg/dL in the liver vein. However, the destruction of the pericentral high affinity detoxification system by glutamine synthetase [15,16] prevents further reduction of ammonia concentrations to normal levels (<30 μg/dL). The reason for the strongly increased ammonia concentrations in portal vein blood (~600 μg/dL) may be due to increased intestinal ammonia production from urea. Because of chronic liver damage mice develop secondary kidney damage [22] with increased blood concentrations of urea. Because of the leaky gut-blood barrier in chronic liver damage, urea reaches bacterial ureases in the intestinal lumen that generate ammonia, which is absorbed into intestinal capillaries and drains into the portal vein [44–46]. However, the latter mechanism still requires experimental validation. The data clearly show that 'periportalization' of liver lobules causes loss of detoxifying functions. The increase in ammonia or urea concentrations to the observed levels will lead to long-term consequences but is not immediately life-threatening. This is in contrast to acute intoxication by hepatotoxicants that may be lethal. In this context, the loss of pericentrally expressed metabolizing enzymes can be interpreted as an adaptive response to chronic CCl4 exposure. CCl4 is metabolically activated by cytochrome P450 2E1 and further CYPs to form the reactive trichloromethyl free radical and the trichloromethyl peroxyl radical [2]. Therefore, the downregulation of cytochrome P450 enzymes is an efficient strategy to survive repeated exposure to toxins that require metabolic activation. This leads to the question as to why also mouse models of periportal fibrosis cause downregulation of pericentral genes. In the present study, we used BDL that is known to preferentially cause periportal liver damage and fibrosis [47,48]. Also, mdr2−/<sup>−</sup> mice are a model of periportal fibrosis [49]. At first glance, it may be difficult to understand why the downregulation of pericentral genes should offer a survival advantage to BDL and mdr2−/<sup>−</sup> mice. In this context, co-evolution between plants and herbivores should be considered [50]. Plants responded to herbivory by formation of plant toxins, many of them metabolically activated by pericentral liver enzymes, such as pyrrolizidine alkaloids and mycotoxins [51]. Herbivores responded by several evolutionary strategies, particularly by novel detoxifying enzymes. Also the periportalization of liver lobules under toxic stress could be interpreted as an adaptive process in animal–plant warfare. Periportalization seems to represent a stereotypical response to different types of inflammatory stress, possibly because the liver lacks the ability to activate distinct adaptive zonation programs for pericentral and periportal damage. This corresponds to previous studies, demonstrating that different types of acute and chronic inflammatory stimuli activate the same gene regulatory networks, whereby upregulation of inflammatory genes occurs simultaneously to the downregulation of metabolic genes [52]; interestingly, both inflammatory and metabolic genes are controlled by the same upstream mechanisms [52]. A strength of the present study is that six individual mice were included for each condition of the time-resolved analysis. This allowed sufficient statistical power to demonstrate the significance of the process of periportalization identified here.

In conclusion, liver fibrosis leads to periportalization of liver lobules. Periportalization occurs as a common response to pericentral and also periportal damage. It allows the liver to adapt to the repeated exposure to hepatotoxic compounds that require metabolic activation by pericentrally expressed enzymes.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4409/8/12/1556/s1: **Supplemental figures.** Figure S1. Immunostaining of CYP2E1 and CPS1 in whole slide scans after BDL. Figure S2. Sirius red and CYP2E1 staining of 8 and 64 week-old mdr2−/<sup>−</sup> mice and 64 week-old wild type (WT) mice. Figure S3. H&E staining of livers (whole slide scans) after 1-year exposure to olive oil and CCl4. Figure S4. H&E staining of livers (whole slide scans) 21 days after BDL compared to sham-operated controls. Figure S5. Overlap of pericentral and periportal genes between three different studies. **Supplemental tables.** Table S1. Significantly (abs(logFC) ≥ 1.5 and FDR ≤ 0.05) up and downregulated genes after 2, 6 and 12 months of treatment of the

*Cells* **2019**, *8*, 1556

mice with 1 g/kg CCl4. Table S2. Characterization of CCl4 induced expression response by PROGENy pathway analysis. Table S3. Transcription factor activities of genes differentially expressed after CCl4 exposure. The column confidence is an empirical classification indicating how confident we are about the TF activity prediction based on the confidence of its regulon. The confidence class range from A–E with being A the highest confident class. Table S4. Analysis of GO group enrichment based on genes differentially expressed after CCl4 exposure. Only those GO terms are shown that were significantly enriched (FDR ≤ 0.05) in at least one-time point. Table S5. Significantly (abs(logFC) ≥ 1.5 and FDR ≤ 0.05) up and downregulated genes after 2, 6 and 12 months of treatment of the mice with olive oil. Table S6. List of zonated pericentral and periportal genes. A. Consensus list including genes that were identified as zonated in at least two of the three studies Braeuning et al., 2006; Saito et al., 2013; Halpern et al., 2017. B. Zonated genes in Braeuning et al., 2006; C. Zonated genes in Saito et al., 2013; D. Zonated genes in Halpern et al., 2017. Table S7. Table of leading-edge genes for pericentral and periportal gene set members that mainly account for the corresponding gene set enrichment score. Table S8. Significantly over-represented gene sets (p-value ≤ 0.05) that characterize the overlap of the CCl4 signature and pericentral/periportal gene set, respectively. **Supplemental Videos.** Supplemental video 1. 3D reconstructions of control mouse liver immunostained for CYP2E1. Supplemental video 2. 3D reconstructions of fibrotic mouse liver on month 6 of repeated CCl4 intoxication immunostained for CYP2E1.

**Author Contributions:** A.G.; J.G.H.: Study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript; obtained funding; study supervision. M.M.; W.M.; R.H.; Y.A.A.; T.A.; E.A.A.; K.M.S.; J.R.; M.-L.B.; M.H.: acquisition of data; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript. C.H.H.; J.S.-R.: Biostatistical analyses of the RNA-seq. data; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript. M.M.-S.; R.G.: critical revision of the manuscript. A.Z.; D.D.: Image analysis; 3D reconstructions; drafting of the manuscript; critical revision of the manuscript. C.T.: Drafting of the manuscript; critical revision of the manuscript; obtained funding.

**Funding:** This study was supported by the BMBF (Germany) funded project LiSyM (FKZ 031L0052; 031L0045; 031L0049).

**Acknowledgments:** We thank Zaynab Hobloss and Brigitte Begher-Tibbe -Leibniz Research Centre for Working Environment and Human Factors, Technical University Dortmund, Dortmund, Germany, for competent technical assistance.

**Availability of Data and Material:** The code to perform the presented RNA-seq study is written in R and is freely available on GitHub: https://github.com/saezlab/LiverPeriportalization [53].

**Conflicts of Interest:** The authors declare no conflicts of interest.
