*3.1. RNA-Seq Demonstrates Downregulation of Pericentral and Upregulation of Periportal Genes in Fibrosis*

Genome-wide expression response caused by CCl4. To study the influence of fibrosis on liver zonation, we established a mouse model with two intraperitoneal injections of 1 g/kg CCl4 per week over 12 months (Figure 1A). Only a relatively mild fibrosis was observed up to six months (Figure 1B). However, between months 6 and 12, the mice progressed into severe fibrosis characterized by wide Sirius red positive fibrotic streets, regenerative nodules and fibrosis-associated macroscopically visible tumor nodules (Figure 1B).

For RNA-seq analysis CCl4-treated mice were processed after 0, 2, 6 and 12 months; olive oil controls were included after 2 and 12 months. Liver tissues of six mice per condition were analyzed (Figure 2A). A principal component analysis (PCA) of the RNA-seq data showed a good clustering of each group of six mice (Figure 2B). Treatment with CCl4 caused a shift in the inverse direction of principal component 1 (PC1) that explains ~30% of the variance in the data (Figure 2B). PC2 represents the combined effects of olive oil, the solvent of CCl4, and aging (Figure 2B). Differential gene expression analysis revealed that 80/85, 95/89 and 261/902 genes were significantly [abs(logFC) ≥ 1.5 and FDR ≤ 0.05] up/downregulated after 2, 6 and 12 months of CCl4 treatment compared to olive oil controls, respectively, with partially very strong, more than 1000-fold expression changes (Figure 2C; lists of differential genes: Table S1). Strongly and consistently upregulated genes (Figure 2D) comprise extracellular matrix-associated genes, such as *Col28a1*, whose role in liver fibrosis is well-known; the variable domains of immunoglobulin heavy chains, suggesting infiltration of B cells/plasma cells [37], e.g., Ighv10-3, Ighv1-9, Ighv4-57-1, Ighv1-22, Ighv4-57-1; the liver-derived peptide hormone hepcidin-2

which supports iron homeostasis [38]; and some factors that so far have not been considered as primary genes affected by liver damage, such as gliomedin (*Gldn*), a protein expressed by myelinating Schwann cells [39] and the leucine-rich repeat and transmembrane domain-containing protein 2 (Lrtm2). Among the strongest downregulated genes (Figure 2D) are several major urinary proteins (Mups), also known as α2μ-globulins, such as Mup19, Mup21, Mup15, Mup17, Mup-ps16, Mup-ps14 and Mup12. Expression of Mup proteins is known to be induced by androgens and they are physiologically relevant, because they bind small hydrophobic molecules, such as steroid hormones, lipids and retinoids in plasma [40]; several cytochrome P450 enzymes; the sushi domain-containing protein 4 (SUSD4) which inhibits complement factors [41]; calpains (e.g., capn11, Capn8) that act as calcium-dependent cysteine proteases, fatty acid elongase 3 (Elovl3); and roquois homeobox protein 1 (lrx1-6) that is known as a cardiac transcription factor [42]. The role of many of these differential genes in liver fibrosis remains unknown. Characterization of the CCl4-induced expression response by pathway analysis using the functional genomics tools PROGENy identified the inflammatory pathways TGFβ, NFκB, TNFα and hypoxia-induced signaling as most active (Figure 2E; Table S2), which is in agreement with previous studies [43]. Estrogen and androgen associated pathways were among the most decreased in the CCl4-exposed livers for all time points compared to corresponding oil samples. Transcription factor (TF) activities were inferred with DoRothEA and were dominated by TF with an increased activity that mediate inflammation (e.g., NFKb1, Stat1) cell stress as well as hypoxia response (Hif1a, Trp53, Atf1) and support proliferation (e.g., E2f1, Ef3, Egr1). TFs with reduced activities are known to mediate mature liver functions, such as Hnf4a, Hnf1a, Esr2 and the Fox genes (Figure 2F; Table S3). Enriched GO-terms such as actin-binding, angiogenesis, cell cycle, death and immune response further round out the picture of an inflamed, regenerating tissue (Figure 2G; Table S4).

**Figure 1.** Mouse model of liver fibrosis induced by CCl4 administration. (**A**) Treatment schedule. (**B**) Macroscopical alterations and visualization of fibrosis by Sirius red staining. Scale bars: 200 μm.

The solvent controls with olive oil alone showed expression changes after 2 and 12 months, respectively, compared to untreated mice at time zero (Table S5). No age-matched untreated controls were included, because the study was designed to identify CCl4 induced expression changes. Administration of CCl4 in oil and comparison to oil controls represents a frequently used protocol.

**Figure 2.** Bioinformatics of RNA-seq data of mouse liver tissue after exposure to CCl4 for up to one year. **A**. Analysis schedule; **B**. Principal component analysis (PCA). Untreated mice at the time point zero (0), the day of onset of exposure for the other mouse groups, period of olive oil exposure in blue (2 and 12 months) and period of CCl4 (solved in olive oil) exposure (2, 6 and 12 months) in red. **C**. Visualization of significantly up (green) and downregulated (blue) genes after 2, 6 and 12 months of CCl4 exposure. **D**. The 20 most up- and downregulated genes after 2, 6 and 12 months exposure to CCl4. **E**. Up- and downregulated pathways via PROGENy. The color legend indicates pathway activity (z-score). **F**. Transcription factor (TF) activities computed with DoRothEA. The color legend indicates TF activity (normalized enrichment score, NES). **G**. Enriched Gene Ontology (GO) groups. The color legend indicates the degree of enrichment (NES).

CCl4-induced expression response in relation to zonated genes. To study a possible zonation of genes up- or downregulated in CCl4-induced fibrosis, a consensus list of pericentral and periportal genes was established, containing 136 and 83 genes, respectively (Table S6). To our knowledge, three previous studies identified genes with zonated expression [31–33], whose overlap was relatively low (Table S6; Figure S5). Hence, genes were included in the consensus list, when they were identified as pericentral or periportal by at least two of the three published studies (Table S6). Genes were ranked by a gene-level statistic (here moderated t-value provided by limma) indicating the strength of their deregulation in response to CCl4 treatment with upregulated genes at the top ranks (left side of the x-axis) and downregulated at the bottom of the list (right) (Figure 3A). Each vertical line on top of the x-axis represents a member of the pericentral or periportal gene set. The y-axis represents the enrichment score (ES), where values higher than zero indicate enrichment of zonated genes among upregulated and values smaller than zero among downregulated genes. Pericentral genes were significantly enriched among downregulated genes at all time points of CCl4 treatment (Gene Set Enrichment Analysis (GSEA), *<sup>p</sup>*-values of 3.99 <sup>×</sup> <sup>10</sup>−<sup>4</sup> for month 2, 3.73 <sup>×</sup> <sup>10</sup>−<sup>4</sup> for month 6 and 2.77 <sup>×</sup> <sup>10</sup>−<sup>4</sup> for month 12). Periportal genes were significantly enriched among upregulated genes only after two and six but not after 12 months of CCl4 treatment (GSEA, *<sup>p</sup>*-values of 1.42 <sup>×</sup> <sup>10</sup>−<sup>4</sup> for month 2, 0.013 for month 6). (Figure 3B). *P*-values were not adjusted for multiple hypothesis testing as we tested only 2 gene sets per signature. Leading edge analysis identified a set of downregulated pericentral and upregulated periportal genes across all time points that are mainly driving the significant GSEA results (Figure 3C, Table S7). We also characterized the overlap of pericentral/periportal genes and the most responsive genes of CCl4 treatment across all time points using over-representation analysis. Analyzing the downregulated pericentral genes, biological processes such as monocarboxylic acid metabolism, epoxygenase P450, and glutamine family catabolic process and the KEGG pathways primary bile acid biosynthesis as well as arginine and proline metabolism were enriched; the transcription factor small heterodimer partner (SHP; synonym: Nr0b2), an interaction partner of HNF4α and LXRα, showed increased activity (Figure 3D). Among the periportal upregulated genes' GO groups associated with lipid metabolism, triglyceride lipase, and phospholipid transport were enriched. Some hits are listed in Figure 3D and the complete list is available in Table S8. Thus, during CCl4-induced liver fibrosis, a complex conglomerate of inflammatory pathways orchestrate downregulation of pericentral and upregulation of periportal genes that further will be referred to as 'periportalized' lobular zonation.

Expression of several genes with a zonated expression pattern was validated by quantitative real-time polymerase chain reaction (qRT-PCR, Figure 4). Analysis of pericentrally expressed genes, solute carrier family 1 member 2 (GLT1), glutamine synthetase (GS), ornithine amino-transferase (Oat), and the vascular/hepatic-type arginine vasopressin receptor (Avpr1a), confirmed a strong downregulation during CCl4 treatment, particularly between months 6 and 12 (Figure 4A). In contrast, the periportal genes glutaminase 2 (Gls2), the urea cycle enzyme carbamoyl phosphate synthetase 1 (CPS1) and the gluconeogenesis enzyme phosphoenolpyruvate carboxykinase 1 (PCK1) showed an increase until month 6, followed by a decrease at month 12 (Figure 4B). The urea cycle enzyme arginase 1 (Arg1) showed little change until month six followed by a moderate decrease at month twelve (Figure 4B). Thus, qRT-PCR of selected genes confirmed a strong decrease of the pericentral genes, while the changes of periportal genes are weaker and more complex, characterized by an increase until month 6 and a decrease between months 6 and 12, an observation that will be interpreted in the context of the immunostaining data described below.

**Figure 3.** Periportalization of CCl4 exposed liver tissue. **A**. Enrichment of pericentral and periportal genes among genes up or downregulated by CCl4 exposure. **B**. Summarized results of Gene Set Enrichment Analysis (GSEA) showing normalized enrichment score (NES) and *p*-values. **C**. Leading edge of periportal and pericentral gene set that mainly accounts for the enrichment score of the gene set. The color scheme indicates the logFC. **D**. Selection of GO-terms and TFs that characterize the overlap of CCl4 signature and pericentral/periportal gene sets.

**Figure 4.** Quantitative real-time polymerase chain reaction (qRT-PCR) confirmation of selected pericentral (**A**) and (**B**) periportal genes. The x-axis represents the time of CCl4 treatment, while the y-axis depicts relative RNA expression normalized to controls (0 months). Glt1: glutamate transporter 1; GS: glutamine synthetase; Oat: ornithine aminotransferase; Avpr1a: arginine vasopressin receptor 1A; Gls2: glutaminase 2; Cps1: carbamoyl phosphate synthetase I; Arg1: arginase 1; Pck1: phosphoenolpyruvate carboxykinase 1. The data are means ± standard errors of 6 mice per time point. \* *p* < 0.05; \*\* *p* < 0.01; \*\*\* *p* <0.001 compared to the untreated controls (0).

Spatio-temporal analysis of periportalization. Further insight into spatio-temporal changes of zonation was obtained by immunostaining. Similar results were observed for the pericentrally expressed enzymes, cytochrome P450 (CYP) 3A, 1A, 2C, 2E1 and GS (Figure 5A). Compared to controls, the CYP positive areas around central veins became narrower after 2 and 6 months of CCl4 administration. However, contacts between CYP2E1 positive areas present in controls (Figure 5B, control; Supplementary Video 1) were maintained even after 6 months of repeated CCl4 treatment, giving the impression of central-to-central bridging (Figure 5B, CCl4; Supplementary Video 2). Until month 12, CYP immunostaining decreased massively. Similarly, GS showed central-to-central bridging at month two and six, followed by an almost complete loss of expression at month 12 (Figure 5A). Image analysis confirmed the decrease of the immunostained CYP1A1-positive area (Figure 5C). In controls, the periportally expressed urea cycle enzymes, arginase 1 and CPS1 showed a periportal to midzonal staining pattern with a relatively narrow negative pericentral zone (Figure 5A). The vessels in the center of arginase 1 or CPS1 positive regions are portal veins, while the vessels in negative regions represent central veins. During the one-year-period of CCl4 administration, fibrotic streets formed between the

central veins, which were particularly obvious between months 6 and 12 (Figure 5A). No expression of arginase 1 or CPS1 occurred in the fibrotic streets, while these enzymes were expressed in hepatocytes at similar levels as in control mice, even in the regenerative nodules at month 12. Whole slide scans immunostained for CYP1A and arginase 1 illustrated the narrowing of the pericentral region expressing CYP1A (months 2 and 4), followed by an almost complete loss at month 12 (Figure 5D). Vice versa, arginase1 expression extended into the pericentral region at months 2, 6 and 12 (Figure 5D).
