Sunday, February 25, 2024
BestWooCommerceThemeBuilttoBoostSales-728x90

Heterogeneity of hepatocyte dynamics restores liver architecture after chemical, physical or viral damage – Nature Communications


Liver homeostasis is driven by proliferation of midlobular hepatocytes

We used three different mouse models to investigate clonal expansion of hepatocytes and their proliferative rates during liver regeneration and homeostasis. We employed multicolor lineage tracing Rosa26-Rainbow (Rosa26rbw) Cre-mediated recombination mice in our study. This model uses pairs of LoxP sequences to randomly recombine 3 of 4 fluorescent proteins (Fig. 1A)17. In the absence of Cre-recombinase-mediated recombination, CAG-EGFP is expressed constitutively resulting in each liver cell permanently expressing green color (Fig. 1A). We crossed these mice with TMX-inducible albumin Cre mice (Alb-CreERT2)18 to generate our research model, Alb-CreERT2 Rosa26rbw mice, to specifically activate the Rainbow colors in cells expressing albumin, a gene differentially expressed in hepatocytes. Injecting the mice with 20 mg/ml TMX (3 consecutive days, 100 μl per day) resulted in hepatocyte-specific recombination throughout the liver lobule. With this approach, hepatocytes were permanently labeled and randomly recombined to express 1 of 3 colors, yellow (mOrange), red (mCherry) and/or light blue (mCerulean) (Fig. 1A). As Cre recombinase expression was restricted to hepatocytes, non-hepatocyte cells including cholangiocytes, stellate cells, immune and endothelial cells remained labeled in green (CAG-EGFP) (Fig. 1A). Hepatocytes only expressed CAG-EGFP and not mOrange, mCherry, or mCerulean without TMX administration, confirming TMX strictly regulated Cre activity (Fig. 1A). 20 mg/ml TMX dose was determined after evaluation of labeling efficiency of both cell hepatocytes and non-hepatocytes in TMX dose-response tests (starting at 0.1 mg/ml up to 20 mg/ml per 1 day) (Supplementary Fig. 1A) and time-course tests (between 1 and 4 consecutive days) (Fig. 1A and Supplementary Fig. 1B). We histologically differentiated central veins from portal tracts due to the identification of CAG-EGFP stained bile ducts around portal veins (Fig. 1A) and CK19 IHC using consecutive slides (Fig. 1B). We confirmed central veins by IHC using consecutive slides for the marker glutamine synthetase (GS) that was expressed exclusively in the first 2–3 layers of hepatocytes surrounding the central veins (Fig. 1B).

Fig. 1: Hepatocyte renewal and ploidy during homeostasis in the healthy mouse liver.

A Schematic showing the 3 mouse models of our study; (Rosa26-Rainbow (Rosa26rbw) Cre-mediated recombination mice, TMX-inducible albumin Cre mice (Alb-CreERT2) and Alb-CreERT2 Rosa26rbw mice) (created with BioRender.com) and representative images of liver sections from 10-week-old male Rosa26rbw and Alb-CreERT2 Rosa26rbw mice (n = 4). The bottom panel shows a detail view of the liver lobule from a Alb-CreERT2 Rosa26rbw male mouse, showing the 3 zones along the portal-central axis; zone 1 or periportal area (PP) that encircles the portal tracts identified by bile ducts (arrowhead); zone 3 or pericentral area (PC) that is located around central veins; and zone 2 or midlobular area (Mid) that is located in between. The Cre-activated fluorophores (CAG-EGFP, mCerulean, mOrange and mCherry) plus DAPI staining are shown in the images. CV central vein, PV portal vein. TMX tamoxifen. Scale bars, 100 μm. B Representative Glutamine Synthetase (GS) and Cytokeratin 19 (CK19) IHC analysis in liver sections from a 10-week-old Alb-CreERT2 Rosa26rbw mouse (n = 4). Scale bar, 100 μm. C Schematic of the heterozygous multicolor Alb-CreERT2 Rosa26rbw mice, that allow us to trace the fate of polyploid or diploid hepatocytes in vivo. Created with BioRender.com. D Total area in μm2 per fluorophore and per photo or field of view (10×) of liver sections from Alb-CreERT2 Rosa26rbw mice. 6–7 photos per mouse; n = 3 mice. Data are presented as mean values +/− SEM. One-way ANOVA, Tukey’s multiple comparisons post-test. E Colored hepatocytes in % per Rainbow fluorophore and per photo or field of view (10×). Data are presented as mean values +/− SEM. 6–7 photos per mouse; n = 3 mice. F Colored hepatocytes in % per Rainbow fluorophore and per area of the liver lobule (PP, Mid and PC). 27 liver lobules analyzed from 3 mice. Data are presented as mean values +/− SEM. One-way ANOVA, Tukey’s multiple comparisons post-test. G Quantification of number of cells per clone (y-axis) and number of clones (x-axis) per area of the liver lobule and per Rainbow fluorophore. Data are presented as mean values. 32 liver lobule areas analyzed from 3 mice (7–8 photos per mouse). One-way ANOVA, Tukey’s multiple comparisons post-test. H Relative frequency in % of each clone size per area of the liver lobule and per Rainbow fluorophore. Data are presented as mean values +/− SEM. 32 liver lobule areas analyzed from 3 mice (7–8 photos per mouse). I Number of total Ki67 positive hepatocytes per area within the liver lobule and per photo or field of view (10×). 48 photos analyzed from 3 mice. Data are presented as mean values +/− SEM. One-way ANOVA, Tukey’s multiple comparisons post-test. Ki67 IHC showing a liver lobule. Gray, red and blue arrows show PP, Mid and PC hepatocytes positive for Ki67. Scale bar, 100 μm. J Hepatocyte ploidy in the liver lobule of 10-week Alb-CreERT2 Rosa26rbw mice. % of total polyploid and diploid hepatocytes within the liver lobule (top left graph) and per area of the liver lobule (top right graph). % of diploid hepatocytes (hepatocytes that express only 1 fluorophore) being mononucleated or multinucleated (bottom left graph). % of polyploid hepatocytes (hepatocytes that express more than 1 fluorophore) being mononucleated or multinucleated (bottom right graph). Data are presented as mean values +/− SEM. Right: Immunofluorescence showing 4 of 5 individual PC hepatocytes expressing more than 1 color. 5 photos from 3 mice were analyzed. Scale bar, 100 μm. Source data are provided as a Source Data file.

One of the main features of the mouse liver is that 90% of hepatocytes are polyploid15. Numerous studies have demonstrated that the ploidy status of hepatocytes defines their proliferation rate and metabolic function5,15,16. Heterozygous multicolor reporter mice have been used to trace the fate of polyploid or diploid hepatocytes in vivo5,19. Given our Alb-Cre Rainbow mouse model was heterozygous, it allowed us to examine hepatocyte ploidy status. In a diploid hepatocyte, one of the two pairs of chromosomes harbors a Rainbow allele, and the cell can express only one fluorophore (Fig. 1C). In contrast, polyploid hepatocytes can be labeled by co-expressing multiple colors (Fig. 1C).

We administrated TMX to heterozygous Alb-Cre Rainbow mice after the polyploidization period of hepatocytes19, at 8 weeks of age. After a 2-week TMX washout period, hepatocytes were stochastically labeled with mOrange, mCherry, or/and mCerulean (Fig. 1A, quantification in 1D). Meanwhile, non-hepatocytes (labeled with CAG-GFP) took up a significantly reduced total area (Fig. 1D). Within the group of hepatocytes, 44% were labeled mCerulean, 26% mCherry and 30% mOrange (Fig. 1E). Next, we calculated the % of colored hepatocytes per zone within the liver lobule, identifying PP hepatocytes, Mid hepatocytes and PC hepatocytes (Fig. 1F). Midlobular hepatocytes expressing mOrange, mCherry or mCerulean were found significantly more numerous than PP or PC, with double the number of hepatocytes compared to the others (Fig. 1F). These results were in line with the number of cells per clone, which were notably more numerous in Mid areas formed mostly by individual hepatocytes followed by clones of 2 and 3 hepatocytes (Fig. 1G). Clones of 4 and 5 hepatocytes were less common and were mostly located in Mid and PC areas, reaching a relatively lower frequency than clone sizes formed by 1–3 hepatocytes (Fig. 1G, H).

Furthermore, we quantified basal levels of hepatocyte proliferation during liver homeostasis, using the proliferative marker Ki67 through IHC on consecutive histological slides. Midlobular hepatocytes showed a higher proliferative rate than PC or PP hepatocytes (Fig. 1I). Our observations agree with recent studies that found modest hepatocyte proliferation in all three zones, with more proliferation occurring in the Mid zone5,8,13, and disagree with the prior concept of streaming of hepatocytes where new hepatocytes are generated in the PP areas and slowly expand toward the central vein20.

As previous studies have found faster cell proliferation rates in diploid rather than polyploid hepatocytes5,15, we wondered if an uneven distribution of ploidy across the liver lobule could account for the increased proliferation in Mid hepatocytes. We, therefore, quantified the ploidy status of hepatocytes per zone of the liver lobule (Fig. 1J) to evaluate whether Mid hepatocytes exhibiting higher Ki67 staining were mostly diploid. We found that 72% of hepatocytes were polyploid and 28% diploid across the lobule in liver homeostasis of mice at 10 weeks of age (Fig. 1J). Within the liver lobule, polyploid hepatocytes were overrepresented in Mid and PC zones making up 76% and 80% respectively compared to PP areas, where 58% of the hepatocytes were polyploid. (Fig. 1J). Even though 81% of diploid hepatocytes were mononucleated, 76% of polyploid hepatocytes were mononucleated as well (Fig. 1J), so nuclear ploidy did not match with the hepatocyte ploidy status in our model. We classified as polyploid hepatocytes those which expressed more than one fluorophore.

Periportal and midlobular hepatocytes repopulate the liver after CCl4-induced acute injury

We next investigated the response of hepatocytes to acute liver injury with CCl4 and their implication in the remodeling of the whole organ mass recovery. CCl4 is a hepatotoxic compound that causes cellular damage to PC hepatocytes as they express the cytochrome P450 enzyme Cyp2e1, which is required to metabolize CCl4 into reactive free radicals21. As shown in Fig. 2A, a single dose of CCl4 was administrated in Alb-CreERT2 Rosa26rbw mice and liver histology was examined at 2 and 12 days after the injection. We found severe liver injury at day 2 after CCl4 treatment by examining liver sections using H&E staining, hepatocyte (mOrange+, mCherry+ or/and mCerulean+ cells) and non-hepatocyte (CAG-EGFP+ cells) immunofluorescence (IF) analysis and GS IHC (Fig. 2B, C). Approximately 2/3 of the hepatocyte mass was depleted, with almost no colored-hepatocytes detected at the PC area and consecutive Mid regions compared to control livers (Fig. 2C, D). H&E staining showed necrotic/apoptotic hepatocytes in PC and Mid regions (Fig. 2B), which correlated with a significant accumulation of TUNEL positive cells (Fig. 2G). GS expression in PC hepatocytes was also reduced at day 2 of CCl4 compared to livers from untreated mice (Fig. 2B). Despite this extensive damage caused by CCl4, half of the Mid hepatocytes and the entire PP hepatocyte population remained morphologically normal (Fig. 2B–C). Notably, we observed fewer individual hepatocytes in these areas comparing to control mice (Fig. 2D and Fig. 1G), predominating clones formed by 2–4 hepatocytes (Fig. 2D) and numerous 5 hepatocyte clones (Fig. 2D). In line with these results, the percentage of the relative frequency of clones formed by 1 hepatocyte was reduced compared to control mice while that of 2–4 hepatocytes increased at day 2 after CCl4 (Figs. 2E and 1H). Clones formed by 5 or more hepatocytes were significantly increased over control mice, with a relative frequency of 5% (Fig. 2E) compared to <1% in livers from control mice (Fig. 1H). Twelve days after the initial CCl4 injection, the liver was completely recovered showing a normal hepatocyte morphology by H&E and IF analysis (Fig. 2B, C). GS positive hepatocytes were nearly restored (Fig. 2B) and dead hepatocytes measured through TUNEL staining were barely detected (Fig. 2G). Clones at PC areas were identified de novo mainly formed by ≥2–4 hepatocytes (Fig. 2D), with a relative frequency of 6–7% (Fig. 2E). Larger clones of 5 hepatocytes were also found although their relative frequency was less abundant than smaller clones and accounted for 4% of the population (Fig. 2D, E). Interestingly, we detected clones formed by 6 hepatocytes only at 12 days after CCl4 treatment. They were concentrated in Mid and PC areas while PP zones mainly were composed of 3–4 hepatocyte clones (Fig. 2D, E). The differences observed in colored-hepatocyte dynamics within the liver lobule at days 2 and 12 after CCl4 suggested that hepatocytes of the 3 zones experienced different proliferation rates to recover lobule integrity. So, we next investigated the location of hepatocytes entering the cell cycle by IHC staining of the proliferative marker Ki67 in consecutive liver samples. Immunostaining for Ki67 showed significantly higher Ki67+ hepatocyte % at day 2 after CCl4-induced damage compared to livers from control mice or mice at day 12 after CCl4 injection (Fig. 2G, I). Notably, we only found proliferating Ki67 hepatocytes in PP and Mid zones with no Ki67+ staining for hepatocytes around central veins (Fig. 2G, I). PP hepatocytes showed a higher proliferative rate than Mid hepatocytes at day 2 after CCl4 injection although it failed to reach the statistical significance (Fig. 2G, I). However, 24 h later, on day 3 after CCl4 treatment, individual PC hepatocytes and some PC hepatocyte clones were generated, as shown in Supplementary Fig. 7A. Three days later, at 6 days after the initial injection with CCl4, larger PC and Mid hepatocyte clones were observed (Supplementary Fig. 7B) and Mid hepatocytes reached their highest proliferative activity at this time point (Supplementary Fig. 7B, C). The number of proliferating Ki67 hepatocytes was drastically reduced in the Mid area after 12 days of CCl4 treatment, with a mean of 0.4 ± 0.16 Ki67+ hepatocytes per field of view while we did not practically find any proliferating Ki67 hepatocyte in PC or PP zones (Fig. 2G, I). These data demonstrate that undamaged Mid hepatocytes repopulate the liver after CCl4-induced acute injury via proliferation giving rise to new hepatocytes which will displace adjacent pre-existing hepatocytes to central vein areas. Displaced hepatocytes acquired their zonated metabolic function as illustrated by GS re-expression on day 12 after CCl4 treatment (Fig. 2B).

Fig. 2: Periportal and midlobular hepatocytes repopulate the liver after CCl4-induced acute injury.
figure 2

A Schematic showing the experimental design for CCl4-induced liver injury. Created with BioRender.com. B Representative H&E and GS staining of liver sections from untreated (control) and CCl4 treated mice (n = 4). Scale bars, 100 μm. C Representative immunofluorescence images of the liver lobule architecture displaying pericentral damage at 2 days post-CCl4 and liver repair at 12 days post-CCl4 (n = 4). Scale bars, 100 μm. D Quantification of number of cells per clone (y-axis) and number of clones (x-axis) per area of the liver lobule and per Rainbow fluorophore. Data are presented as mean values. 11–13 liver lobule areas were analyzed per condition (3–7 photos per mouse); 3 mice per group. One-way ANOVA, Tukey’s multiple comparisons post-test. E Relative frequency of each clone size per area of the liver lobule and per Rainbow fluorophore. Data are presented as mean values +/− SEM. 11–13 liver lobule areas were analyzed per condition (3–7 photos per mouse); 3 mice per condition. F Quantification of the % positive area per field of view (10×) of CAG-EGFP in liver sections from untreated and CCl4 treated mice. Data are presented as mean values +/− SEM. 8 photos per mouse; 3 mice per group. One-way ANOVA, Tukey’s multiple comparisons post-test. G Ki67 IHC and TUNEL assay in mouse liver sections at 2 and 12 days after CCl4. Ki67: red arrows indicate non-hepatocyte cells positive for Ki67 while black arrows show proliferating Ki67 hepatocytes. TUNEL: black arrows indicate dead hepatocytes. Scale bars, 100 μm. H CD68 and CD45 IHC of liver sections from mice at 2 and 12 days after CCl4. Scale bars, 100 μm. I Quantification of the Ki67 staining. Left graph: Total Ki67 positive cells (hepatocytes and non-hepatocyte cells) in % per field of view (10×) of liver sections from mice at 2 and 12 days after CCl4 treatment. 27, 12 and 23 photos from 3 mice for the untreated, 2 days post-CCl4 and 12 days post-CCl4 groups, respectively. Middle graph: proliferating Ki67 hepatocytes per field of view (10×) in each zone of the liver lobule (PP, Mid and PC) after 2 days of CCl4 exposure. 12 photos from 3 mice. Right graph: proliferating Ki67 hepatocytes per field of view (10×) in each zone of the liver lobule after 12 days of CCl4. 23 photos from 3 mice. Data in all panels are presented as mean values +/− SEM. One-way ANOVA, Tukey’s multiple comparisons post-test. J % Apoptosis+ area from the images shown in section G. Data are presented as mean values +/− SEM. 7–8 photos per mouse; n = 3 mice per group. One-way ANOVA, Tukey’s multiple comparisons post-test. K Quantification of the CD68 and CD45 IHC. CD68 and CD45 area in % per field of view (10×) in liver sections from untreated control and CCl4 treated mice. CD68: 10–18 photos were analyzed per mouse; n = 3 mice per group. Data are presented as mean values +/− SEM. CD:45: 7–18 photos per mouse; n = 3 mice per group. Data are presented as mean values +/− SEM. One-way ANOVA, Tukey’s multiple comparisons post-test. L Analysis of hepatocyte ploidy in the liver lobule from mice after 12 days recovery post-CCl4 administration. % of total polyploid and diploid hepatocytes within the liver lobule (left graph) and per area of the liver lobule (middle left graph). % of diploid hepatocytes (hepatocytes that express only 1 fluorophore) being mononucleated or multinucleated (middle right graph). % of polyploid hepatocytes (hepatocytes that express more than 1 fluorophore) being mononucleated or multinucleated (right graph). 5 photos per mouse; n = 3 mice. Data are presented as mean values +/− SEM. Source data are provided as a Source Data file.

The proportion of non-hepatocytes, tracked by CAG-EGFP staining was significantly increased at day 2 after CCl4 treatment compared to control mice and mice sacrificed after the recovery period at day 12 after the injury (Fig. 2F). Since CCl4 toxicity causes macrophages and monocytes infiltration which facilitate the removal of cellular debris generated by necrosis and apoptosis of hepatocytes, we next examined leukocyte and macrophage infiltration during liver repair. To this end, we performed IHC analysis of CD68+ (macrophage marker) and CD45+ (leukocyte marker) cells at early (2 days) and late stage (12 days) of liver regeneration after CCl4 -induced damage. We observed a marked accumulation of CD68+ macrophages and CD45+ cells around central veins 2 days after CCl4 injection (Fig. 2H, K). Ten days later, macrophage area was markedly reduced although the influx of CD45 immune cells was still significantly elevated with respect to control mice (Fig. 2H, K). Finally, we quantified the ploidy status of hepatocytes in each zone of the liver lobule after 12 days of CCl4 treatment to evaluate whether hepatocytes generated de novo in PC and Mid areas during the recovery phase differed in ploidy from control mice (Fig. 2L, Fig. 1J). We found that 76% of hepatocytes are polyploid and 24% diploid within the liver lobule (Fig. 2L), similar values to those found in liver homeostasis (Fig. 1J). Like in control mice, Mid and PC zones resulted in having a majority of polyploid hepatocytes, content with a 79% and 86% respectively compared to PP areas, where 61% of the hepatocytes were polyploid. However, PP was mostly diploid (39%), followed by Mid (21%) and PC (14%) (Fig. 2L). Although 81% of diploid hepatocytes were mononucleated, 86% of polyploid hepatocytes turned out to also be mononucleated (Fig. 2L). So, we confirm with this model of CCl4 damage the previous findings in healthy livers: that nuclear ploidy does not match hepatocyte ploidy (Fig. 1J).

Midlobular hepatocytes remodel the liver architecture after CCl4-induced chronic injury

We next analyzed liver lobule changes in response to chronic CCl4 lesion over time (Fig. 3). Ten-week-old Alb-CreERT2 Rosa26rbw mice received 12 injections of CCl4 or corn oil every 2 days for 4 weeks and then they were sacrificed at weeks 14, 16 and 20 of age (Fig. 3A). Mice that received CCl4 treatment experienced a gradual elevation in their body weight over the course of additional injections with CCl4 (Supplementary Fig. 3A). After the sacrifice, we also observed a positive trend in the spleen weight in CCl4-treated animals, compared to control mice, immediately after CCl4 treatment cessation, at week 14, but they did not reach statistical significance (Supplementary Fig. 3B). Nevertheless, liver weight was remarkably increased in CCl4-treated mice at week 14, followed by week 16, whereas, at week 20, liver weight from control and CCl4 mice had similar values (Supplementary Fig. 3B). As it is shown by H&E and Sirius red staining, additional doses of CCl4 caused chronic injury, that was confirmed by significant pericentral fibrosis in mice sacrificed at week 14 (Fig. 3B, E). Fibrosis area decreased throughout the following two time points, with week 16 showing significant reduction and week 20 nearing the values of livers from control mice (Fig. 3B, E and Supplementary Fig. 3C). These data were consistent with the non-hepatocyte CAG-EGFP staining, with a positive area significantly increased at week 14 but markedly reduced over time during the recovery periods at weeks 16 and 20 (Fig. 3C, G). Our IHC staining also revealed that GS hepatic expression also was changed upon CCl4 exposure since PC hepatocytes reduced the expression of this enzyme at week 14 but this was entirely recovered during the repair period at week 20 (Fig. 3B), with a similar expression pattern to that found in livers from control mice (Fig. 1B). However, the expression of the cholangiocyte marker CK19 showed no alteration at week 14 after CCl4 repeated exposure and at weeks 16 and 20 during the recovery periods (Supplementary Fig. 3D). Chronic CCl4 damage promoted the development of large hepatocyte clones throughout the liver lobule compared to control mice (Fig. 3C, H and Supplementary Fig. 3E, F). After 12 doses of CCl4, we found large clones formed by 8–12 hepatocytes along the 3 zones of the liver lobule in mice sacrificed at week 14 (Fig. 3C, H). At week 20, larger-sized clones formed by 24–16 hepatocytes covered the entire distance from the central to portal veins (Fig. 3C, D, H). We also observed large clones that were extended from the portal triad areas to each other and some isolated medium-sized clones in the Mid zones (Fig. 3D). Clones formed by more than 10 hepatocytes were predominated compared to control mice (Supplementary Fig. 3F), with a relative frequency of 5% (Fig. 3F) versus 0.2% (Supplementary Fig. 3F). Analysis of hepatocyte area at week 20 also showed that large hepatocytes were found within the clones in all 3 different areas of the liver lobule, suggesting that they also exhibit mechanisms of hypertrophy in addition to proliferative events (Fig. 3I, Supplementary Fig. 9). In terms of hepatocyte ploidy status, 66% of the hepatocyte population generated were diploid while 34% of hepatocytes were polyploid (Fig. 3P). These data are significantly different from the hepatocyte ploidy found during liver homeostasis in the healthy liver (Fig. 1J), or liver regeneration in mice treated with only one dose of CCl4 (Fig. 2L). These displayed around 70% polyploid and 30% diploid hepatocytes (Fig. 2L). At week 20, we also found differences in hepatocyte ploidy within the 3 zones of the liver lobule, with more numerous diploid hepatocytes in PP (84%) and Mid areas (71%) and fewer polyploid hepatocytes in the PC area (39%) (Fig. 3P). Although 93% of diploid hepatocytes were mononucleated, 90% of polyploid hepatocytes were mononucleated as well (Supplementary Fig. 3H), similar to values observed in livers from control mice (Fig. 1J) and those that received acute CCl4 damage (Fig. 2L).

Fig. 3: Midlobular hepatocytes remodel the liver architecture after CCl4-induced chronic injury.
figure 3

A Schematic showing the experimental design for the CCl4-induced chronic liver injury model. Created with BioRender.com. B Liver sections stained with H&E, Sirius red and GS from mice at week 14, 16 and 20 of age after receiving additional doses of CCl4. Scale bars, 100 μm. C Representative immunofluorescence images of liver sections from mice at 14, 16 and 20 weeks of age after CCl4 treatment (n = 4). Scale bars, 100 μm. D Detailed view of the liver lobule architecture showing the large hepatocyte clones generated during the recovery period in mice at week 20 of age, 6 weeks post-CCl4 chronic damage. Black arrows show clonal expansion originated from PP hepatocytes to PC zones. Portal tracts are identified by bile ducts (arrowhead). Red circles show isolated midlobular clones. Scale bars, 100 μm. E Quantification of Sirius red staining in liver sections from mice treated with vehicle or CCl4 for 4 weeks (shown in Fig. 3B) and euthanized at week 14, 16 and 20 of age. Data are presented as mean values +/− SEM. 8–5 photos per mouse from a total of 3 mice per group. One-way ANOVA, Tukey’s multiple comparisons post-test. F Relative frequency in % of each clone size per area of the liver lobule and per Rainbow fluorophore from mice at week 20 of age, 6 weeks post-CCl4 chronic damage. Data are presented as mean values +/− SEM. 46 liver lobule areas analyzed from a total of 3 mice (5 photos per-mouse). G Quantification of the % for the positive area per field of view (10×) for CAG-EGFP in liver sections from vehicle and CCl4 treated mice euthanized at week 14, 16 and 20 of age. Data are presented as mean values +/− SEM. 8 photos analyzed per mouse from 3 mice per group. One-way ANOVA, Tukey’s multiple comparisons post-test. H Quantification of number of cells per clone (y-axis) and number of clones (x-axis) per area of the liver lobule and per Rainbow fluorophore from mice treated with CCl4 for 4 weeks and euthanized at week 14 and 20 of age. Data are presented as mean values. 31–46 liver lobule areas analyzed from 3 mice per group (4–5 photos per-mouse). One-way ANOVA, Tukey’s multiple comparisons post-test. Right image: Immunofluorescence staining from mice at week 20 of age showing a large hepatocyte clone formed by 11 cells. 10 cells expressed only one color indicating that most cells of the clones generated during the recovery period after CCl4 were diploid. Scale bars, 100 μm. I Hepatocyte size of clones developed within the liver lobule during the recovery period at 20 weeks of age from mice chronically treated with vehicle or CCl4. Data are presented as mean values +/− SEM. 52–63 liver lobule areas analyzed from 3 mice per condition (5 photos per mouse). One-way ANOVA, Tukey’s multiple comparisons post-test. J Images of Ki67, CD45 and CD68 IHC and TUNEL assay in liver sections from mice treated with CCl4 for 4 weeks and euthanized at week 14. Scale bars, 100 μm. K Quantification of total Ki67 positive cells (hepatocytes and non-hepatocyte cells) in % per field of view (10×) of liver sections from mice treated with vehicle or CCl4 for 4 weeks and euthanized at week 14, 16 and 20 of age. Data are presented as mean values +/− SEM. 23–40 photos analyzed from 3 mice per group. One-way ANOVA, Tukey’s multiple comparisons post-test. L Quantification of proliferating Ki67 hepatocytes per field of view (10x) in each zone of the liver lobule from mice treated with CCl4 for 4 weeks and euthanized at week 14. Data are presented as mean values +/− SEM. 18–20 photos analyzed from 3 mice. One-way ANOVA, Tukey’s multiple comparisons post-test. M Quantification of TUNEL staining. Apoptosis area in % in liver sections from untreated control and CCl4 treated mice euthanized at weeks 14, 16 and 20 of age. Data are presented as mean values +/− SEM. 7–12 photos analyzed per mouse: 3 mice per group. One-way ANOVA, Tukey’s multiple comparisons post-test. N Quantification of CD45 IHC. CD45 area in % per field of view (10×) in liver sections from untreated control and CCl4 treated mice euthanized at weeks 14, 16 and 20 of age. Data are presented as mean values +/− SEM. 10–21 photos analyzed per mouse: 3 mice per group. One-way ANOVA, Tukey’s multiple comparisons post-test. O Quantification of CD68 IHC. CD68 area in % per field of view (10×) in liver sections from untreated control and CCl4 treated mice euthanized at weeks 14, 16 and 20 of age. Data are presented as mean values +/− SEM. 8–15 photos analyzed per mouse: 3 mice per group. One-way ANOVA, Tukey’s multiple comparisons post-test. P Hepatocyte ploidy in the liver lobule of mice at 20 weeks of age or 6 weeks post-CCl4 chronic damage. % of total polyploid and diploid hepatocytes within the liver lobule (left graph) and per area of the liver lobule (right graph). Data are presented as mean values +/− SEM. 5 photos analyzed per mouse; n = 3 mice. Source data are provided as a Source Data file.

We next examined the proliferation of hepatocytes after extended CCl4 treatment. Immunostaining for Ki67 showed a significantly higher percentage of proliferating Ki67+ hepatocytes at week 14 after additional doses of CCl4 compared to control mice (Fig. 3J, K). Mainly, we found significantly more Ki67+ hepatocytes per field of view in the Mid and PC areas than in the PP region at this time point (Fig. 3L). However, we did not identify any proliferating Ki67+ hepatocytes during the recovery times at weeks 16 and 20 or after 2- and 6-weeks post damage respectively (Fig. 3K and Supplementary Fig. 3G). We also quantified liver apoptosis levels through TUNEL staining after chronic treatment with CCl4 (Fig. 3M). We found that most TUNEL positive cells were hepatocytes around central veins and adjacent to Mid areas by IHC of consecutive slides at week 14 (Fig. 3J). Moreover, we also observed apoptotic non-hepatocyte cells around portal triads and central veins (Fig. 3J). In both cell-types, TUNEL staining was scarce although statistically significant compared to livers from control mice (TUNEL staining at week 14) (Fig. 3J, M and Supplementary Fig. 3G). Apoptosis was notably reduced at the recovery time points at weeks 16 and 20, similar to values of livers from control mice (Fig. 3M). Given we detected a significant elevation of proliferating Ki67+ non-hepatocyte cells after the 12 doses of CCl4 in week 14 (Fig. 3K), we further examined the contribution of liver immune cells to this increase. We quantified CD45 and CD68 positive cells by IHC (Fig. 3J, N, O). Overall, we identified a considerable concentration of CD45+ cells around the central veins at week 14 (Fig. 3J), which rapidly dropped during the recovery periods of weeks 16 and 20, like those counted in livers from control mice (Fig. 3N and Supplementary Fig. 3G). The pattern for CD68+ cells was similar to that of CD45+ staining but it was less intense (Fig. 3J, O and Supplementary Fig. 3G).

DDC injury-induced ductular reaction causes proliferation of midlobular hepatocytes

We next used the DDC diet to induce both biliary and hepatic lineage damages and evaluate potential mechanisms of differentiation, dedifferentiation and transdifferentiation in vivo. To track the liver cells, we injected 8-week-old Alb-CreERT2 Rosa26rbw mice with TMX and, after a 2-week washout period we fed the mice with a chow or a 0.1% (w/w) DDC diet for 6 weeks (Fig. 4A). We then sacrificed the mice and measured body, liver and spleen weights (Fig. 4C). Although the DDC diet was well tolerated, mice exhibited a significant reduction in their body weights compared to mice fed a chow diet (Fig. 4C). DDC treatment did not alter the spleen weight (Fig. 4C); however, it did cause significant hepatomegaly illustrated by a notable increase in liver weight (Fig. 4C). To deeply characterize this type of lesion, we histologically analyzed the liver tissues from both diet interventions (Fig. 4B–N). As previously described, after the DDC treatment, mice developed cholangitis evidenced by the presence of biliary porphyrin deposition and periductular fibrosis in H&E and Sirius Red staining respectively (Fig. 4B, G). Immunofluorescence analysis revealed that the DDC diet caused the death of almost all PP hepatocyte population (Fig. 4G). However, some individual hepatocytes surrounded by the bile ducts preserved cytoplasmatic Rainbow fluorescence, and normal non-damaged nucleus (Fig. 4G). Increased cellular apoptosis was validated by the increased positive TUNEL staining around portal veins in these mice (Fig. 4G, I, L). However, no apoptotic cells were found in livers from control mice (Fig. 4G, I, L). In addition, liver sections of these mice showed a significant elevation in the number of total cells represented by DAPI staining (Fig. 4D) and % of the non-hepatocyte area as the CAG-EGFP staining exhibited compared to control mice (Fig. 4E). We confirmed by IHC of consecutive sections that a prominent CAG-EGFP+ cell population corresponded to biliary cells using the ductal marker CK19 (Fig. 4F). CK19 staining was overexpressed in liver sections from DDC treated mice for cells whose cytoplasm was smaller than those from hepatocytes confirming that portal triads were expanded with numerous bile ducts of poorly defined lumens (Fig. 4F). CK19 staining was exclusively expressed in cholangiocytes or biliary cells (Fig. 4F). We did not observe any hepatocyte-like cell morphology positive for the marker of non-hepatocytes CAG-EGFP (Fig. 4G). We did not see either any hepatocyte positive for both, any rainbow color (mCherry+, mOrange+ y/o mCerulean+) and CK19 and none of the hepatocytes were arranged in a ductular configuration (Fig. 4G, F). Furthermore, our immunofluorescence analysis did not reveal any Rainbow colored biliary-like cell (Fig. 4G). Taken together, these results confirm that there is not any mechanism of transdifferentiation between biliary and hepatic epithelia as we hypothetically could have found in our multicolor lineage tracing Alb-CreERT2 Rosa26rbw mice (Supplementary Fig. 4A, B). In control mice, CK19 positive staining was restricted to the cells that form the bile ducts of the portal triads (Fig. 4F).

Fig. 4: DDC injury-induced ductular reaction causes proliferation of midlobular hepatocytes.
figure 4

A Schematic showing the experimental design for the DDC-induced bile duct and hepatocyte injury model. Created with BioRender.com. B Images and quantification of Sirius red staining in liver sections from mice fed chow (control mice) or DDC diet for 6 weeks. Data are presented as mean values +/− SEM. 11 photos from 3 mice per group. Unpaired t test. CV central vein, PV portal vein. Scale bars, 100 μm. C Body weight (g), liver weight (mg) and spleen weight (mg) from mice fed chow or DDC diet for 6 weeks; Data are presented as mean values +/− SEM; n = 4 mice per diet. Unpaired t test. D Quantification of total number of cells by DAPI staining in mice fed chow or DDC diet for 6 weeks. Data are presented as mean values +/− SEM. 9 photos analyzed from 3 mice per group. Unpaired t test. E Quantification of the % for the positive area per field of view (10×) for CAG-EGFP in liver sections from mice fed chow and DDC diet for 6 weeks. Data are presented as mean values +/− SEM. 5–14 photos analyzed per mouse; n = 3 mice per group. Unpaired t test. F Representative CK19 and GS IHC analysis in liver sections from mice fed chow or DDC diet for 6 weeks (n = 4). Scale bars, 100 μm. G Liver sections from mice fed chow (left panels) or DDC diet (right panels) for 6 weeks. Top panels show CAG-EGFP and DAPI staining. Middle panels indicate the Rainbow fluorophores. Lower panels represent H&E staining showing a portal vein from both groups of mice. Asterisks indicate biliary porphyrin deposition. Scale bars, 100 μm. H Top graphs: quantification of number of cells per clone (y-axis) and number of clones (x-axis) per area of the liver lobule and per Rainbow fluorophore of liver sections from mice fed chow (control) or DDC diet for 6 weeks. Data are presented as mean values. 22–50 liver lobule areas analyzed from 3 mice per group (7–10 photos per mouse). One-way ANOVA, Tukey’s multiple comparisons post-test. Bottom graphs: relative frequency in % of each clone size per area of the liver lobule and per Rainbow fluorophore from mice fed chow (control) or DDC diet for 6 weeks. Data are presented as mean values +/− SEM. 22–50 liver lobule areas analyzed from 3 mice per group (7–10 photos per-mouse). I Ki67 IHC and TUNEL assay in liver sections from mice fed chow (control) or DDC diet for 6 weeks. Scale bars, 100 μm. J CD68 and CD45 IHC in liver sections from mice fed chow (control) or DDC diet for 6 weeks. Scale bars, 100 μm. K Quantification of the Ki67 IHC. Left graph: Total Ki67 positive cells (hepatocytes and non-hepatocyte cells) in % per field of view (10×) of liver sections from mice fed chow (control) or DDC diet for 6 weeks. 20 and 30 photos analyzed from 3 mice for control and DDC mice respectively. Right graph: proliferating Ki67 hepatocytes per field of view (10×) in each zone of the liver lobule from mice fed chow (control) or DDC diet for 6 weeks. 20 and 30 photos analyzed from 3 mice for control and DDC mice respectively. Data in left and right panels are presented as mean values +/− SEM. One-way ANOVA, Tukey’s multiple comparisons post-test. L Quantification of the TUNEL positive staining. Apoptosis area in % in liver sections from mice fed chow (control) or DDC diet for 6 weeks. 7–18 photos analyzed per mouse: 3 mice per group. Data are presented as mean values +/− SEM. Unpaired t test. M Quantification of the CD68 and CD45 positive staining. CD68 and CD45 area in % per field of view (10×) in liver sections from mice fed chow (control) or DDC diet for 6 weeks. CD68: Data are presented as mean values +/− SEM. 12–13 photos analyzed per mouse; 3 mice per group. Unpaired t test. CD45: Data are presented as mean values +/− SEM. 12–14 photos analyzed per mouse; 3 mice per group. Unpaired t test. N Hepatocyte ploidy in the liver lobule of mice fed DDC diet for 6 weeks. % of total polyploid and diploid hepatocytes within the liver lobule (left graph) and per area of the liver lobule (middle left graph). % of diploid hepatocytes (hepatocytes that express only 1 fluorophore) being mononucleated or multinucleated (middle right graph). % of polyploid hepatocytes (hepatocytes that express more than 1 fluorophore) being mononucleated or multinucleated (right graph). Data are presented as mean values +/− SEM. 5 photos analyzed per mouse; 3 mice. Source data are provided as a Source Data file.

In concurrence with previous work, these data confirmed that our model of 6-weeks of DDC injury stimulated the proliferation of CK19 positive biliary epithelial cells, forming multiple bile ducts in the portal triads, an event known as ductular reaction (DR). Given DR is associated with inflammatory cell infiltration in portal areas and the presence of numerous CK19 negative non-hepatocyte cells by CAG-EGFP, we investigated the expression of CD45 and CD68 markers after DDC feeding (Fig. 4J, M). We observed CD45+ cells significantly accumulated in PP regions (Fig. 4J). This population could correspond with infiltrating lymphocytes that collaborate to the local inflammatory response induced by biliary obstruction. However, we did not observe notable differences in CD68 staining between DDC treated and untreated mice (Fig. 4J, M). In addition, we readily detected clones containing 4–9 hepatocytes after DDC exposure through our immunofluorescence analysis (Fig. 4G, H). These clones were primarily located in the Mid areas of the liver lobule (Fig. 4G, H) and their cell and nuclear sizes were smaller than those from PC hepatocytes. Liver sections from control mice displayed clones formed by 1–3 hepatocytes (Fig. 4G, H), as previously observed in liver homeostasis (Fig. 1G, H). In agreement with this data, liver lobules from mice treated with the DDC diet showed a dramatic increase in Ki67 Mid hepatocyte staining compared to their PP or PC neighbor populations (Fig. 4I, K). GS expression, considered a marker of PC hepatocytes, was reduced in mice fed a DDC diet, with the most GS positive staining concentrated, in the first layer of PC hepatocytes (Fig. 4F). Meanwhile, livers from control mice showed an expanded GS expression though 1–3 layers of hepatocytes (Fig. 4F). These observations suggest DDC injury also damaged the functionality of PC hepatocytes, reducing the expression of key metabolic enzymes, although not their proliferative potential, given we observed some Ki67 positive staining in PC hepatocytes (Fig. 4I). Finally, we characterized the hepatic ploidy in each zone of the liver lobule after 6 weeks of DDC diet to evaluate whether new hepatocytes generated in Mid areas differed in ploidy from control mice (Figs. 4N and 1J). We found 74% polyploid and 26% diploid hepatocytes within the liver lobule (Fig. 4N), similar values to those found in liver homeostasis (Fig. 1J). Like in control mice, Mid and PC zones resulted in having a majority of polyploid hepatocytes, content with a 71% and 88% respectively compared to PP areas, where 65% of the hepatocytes were polyploid. However, in this damage model, PP was mostly diploid (35%), followed by Mid (29%) and PC (12%) (Fig. 4N). Although 78% of diploid hepatocytes were mononucleated, 81% of polyploid hepatocytes turned out to also be mononucleated (Fig. 4N). So, with this model of cholestatic liver damage we confirm our previous findings in healthy and CCl4 injury livers: that nuclear ploidy does not match hepatocyte ploidy (Fig. 4N).

Midlobular hepatocytes repopulate the liver during regeneration after 2/3 PHx

We also explored the remarkable ability of the liver to regenerate itself by examining the contribution of different hepatocyte populations in restoring liver lobule mass. To this end, we subjected TMX-treated Alb-CreERT2 Rosa26rbw mice to 2/3 PHx and we traced hepatocyte clonal dynamics after physical injury. Mice were sacrificed at 24, 48, 72 and 162 h (or 7 days) after PHx (n = 2 per group) which encompassed the main regenerative burst of PHx (Fig. 5A). The removed liver lobes were used as hour 0 controls. Weights from the whole body, liver and spleen, as well as liver/body weight ratio were measured for each time point (Fig. 5B). As expected, liver weight decreased at 24 h, increased at 48 h and almost regained its initial weight at 72 h after resection. After this latter time point, liver weight remained constant until 162 h (Fig. 5B). Mouse body weight dropped at 24 and 48 h after PHx. From this point on, body weight began to increase with a slight recovery at 72 h maintained until 162 h post-resection. The fast rebound in liver weight and continued loss of body weight at 48 h culminated in rapid restoration of the liver/body weight ratio, which was almost completely reestablished by 48–72 h. This ratio was restored even though liver and whole-body weights did not reach their initial values within the time points studied (Fig. 5B). Following PHx, spleen weight gradually increased over time (Fig. 5B). We also investigated the clonal expansion of hepatocytes in the regenerating liver (Fig. 5C–E). At time 0 h, most of the hepatocytes were individual cells or clones formed by 2 or 3 hepatocytes although a few clones formed by 4 hepatocytes were also observed (Supplementary Fig. 5A). These clones were more numerous in the Mid area compared to PP or PC areas (Supplementary Fig. 5A). These findings were similar to those which we observed previously during liver homeostasis (Fig. 1G, H). After 24 h of PHx, individual hepatocytes were less common that at 0 h (Fig. 5C–E). The decrease in individual hepatocytes was concurrent with an increase in the relative frequency of clones formed by 2, 3 and 4 hepatocytes, principally this last one (Fig. 5C–E). Midlobular zones were where we mainly observed more numerous and larger clones (Fig. 5C–E). This clonal expansion continued through 48 h after the resection when the number and size of clones increased considerably, appearing clones of 5 hepatocytes. This dynamic was decreased steadily over the 72 and 162 h time points (Fig. 5C–E). Similarly, to previous time points, Mid hepatocytes were more likely to be larger and more numerous than PP or PC hepatocytes at 48, 72 and 162 h after the injury (Fig. 5C, D). We also examined hepatocyte proliferation by IHC for Ki67 on consecutive liver sections (Fig. 5G, Supplementary Fig. 5A,C). Our results showed Ki67 proliferating hepatocytes were significantly increased at 24 and 48 h after the injury (Fig. 5G and Supplementary Fig. 5C). This maintained hepatocyte proliferative wave tapered down through the 72 and 162 time points (Fig. 5G and Supplementary Fig. 5C). Proliferating hepatocytes were reduced at 72 h, followed by 162 h (Fig. 5G and Supplementary Fig. 5C). We found that Mid hepatocytes were the most proliferative at 24, 48 and 72 h after PHx (Fig. 5H). Taken together, these results indicate that Mid hepatocytes were the primary source of hepatocyte progeny in the regenerating liver. Beyond hepatocyte hyperplasia, we also investigated hypertrophy measuring hepatocyte size during liver regeneration. We notably found increased hepatocyte size at 48 h after the injury compared to time point 0 h (Fig. 5I). The largest hepatocytes were in the Mid region, followed by the PC zone (Fig. 5I).

Fig. 5: Midlobular hepatocytes repopulate the liver during regeneration after 2/3 PHx.
figure 5

A Schematic showing the experimental design for the evaluation of liver regeneration after a 2/3 partial hepatectomy model. Created with BioRender.com. B Liver (mg), spleen (mg) and body weights (g) and liver/body weight ratio (%) at time points 0, 24, 48, 72 and 162 h of mice sujected to 2/3 PHx. N = 2 mice used per time point. Data are presented as mean values. C Liver sections from mice subjected to 2/3 PHx after 24, 48, 72 and 162 h. Scale bars, 100 μm. D Quantification of number of cells per clone (y-axis) and number of clones (x-axis) per area of the liver lobule and per Rainbow fluorophore from mice subjected to 2/3 PHx after 24, 48, 72 and 162 h. Data are presented as mean values. 11–12 liver lobule areas analyzed from 2 mice per group (4 photos per-mouse). One-way ANOVA, Tukey’s multiple comparisons post-test. E Relative frequency in % of each clone size per area of the liver lobule and per Rainbow fluorophore from mice subjected to 2/3 PHx after 24, 48, 72 and 162 h. Data are presented as mean values. 11–12 liver lobule areas analyzed from 2 mice per group (4 photos per-mouse). F Quantification of the % for the positive area per field of view (10×) for CAG-EGFP in liver sections from mice subjected to 2/3 PHx at time points 0, 24, 48, 72 and 162 h. Data are presented as mean values +/− SEM. 7–18 photos per mouse; 2 mice per group. One-way ANOVA, Tukey’s multiple comparisons post-test. G Ki67 IHC in liver sections from mice subjected to 2/3 PHx after 24, 48, 72 and 162 h. Arrows indicate proliferating Ki67 hepatocytes. N = 2 mice. Scale bars, 100 μm. H Quantification of proliferating Ki67 hepatocytes per field of view (10×) in each zone of the liver lobule from mice subjected to 2/3 PHx after 24, 48 and 72 h. Data are presented as mean values +/− SEM. 10–15 liver lobule areas analyzed from 2 mice per group. One-way ANOVA, Tukey’s multiple comparisons post-test. I Left: Hepatocyte size during liver regeneration in liver sections from mice subjected to 2/3 PHx at time points 0, 24, 48, 72 and 162 h. 9–12 liver lobule areas analyzed from 2 mice. Right: Hepatocyte size within the liver lobule from mice 48 h after being subjected to 2/3 PHx. 9 liver lobule areas analyzed from 2 mice. Data in left and right panels are presented as mean values +/− SEM. One-way ANOVA, Tukey’s multiple comparisons post-test. J Images after 72 h of PHx and quantification of % CAG-EGFP positive hepatocyte cells-like of liver sections from mice subjected to 2/3 PHx. Arrowheads indicate hepatocyte cell-like positive for both fluorophores, the non-hepatocye marker CAG-EGFP and the hepatocyte marker mOrange. Arrows indicate hepatocyte cell-like positive for the non-hepatocye marker CAG-EGFP. Dashed boxes show a detail view of a hepatocyte cell-like (arrow) positive for CAG-EGFP, mCherry and mCerulean. Lower left graph represents the % of CAG-EGFP positive hepatocyte cells-like per field of view (10×) of liver sections from mice subjected to 2/3 PHx at time points 0, 24, 48, 72 and 162 h. Data are presented as mean values +/− SEM. Scale bars, 100 μm. K Hepatocyte ploidy in the regenerated liver from mice after 162 h of being subjected to 2/3 PHx. % of total polyploid and diploid hepatocytes within the liver lobule (left graph) and per area of the liver lobule (right graph). Data are presented as mean values +/− SEM. 5 photos analyzed per mouse, 2 mice per group. Source data are provided as a Source Data file.

Quantitative analysis of CAG-EGFP non-hepatocyte cells spiked at 48 h after the injury, followed by a second peak at 72 h (Fig. 5F). As liver regeneration is remarkably accompanied by intrahepatic angiogenesis3, we quantified the marker of endothelial cells CD31 by IHC to evaluate tissue repair after PHx (Supplementary Fig. 5D). We observed an enrichment of CD31+ cells in the regenerating liver during this interval, being in line with our CAG-EGFP staining (Fig. 5F and Supplementary Fig. 5D). Intriguingly, CAG-EGFP immunofluorescence showed some hepatocyte-like cells during liver regeneration at 24, 48, 72 and 162 h after PHx (Fig. 5J and Supplementary Fig. 5E). This hepatocyte-like cell group was first observed at 24 h and increased in number at 48 h and 72 h (Fig. 5J and Supplementary Fig. 5E). Primarily located in the Mid area of the regenerating liver lobules, this CAG-EGFP+ cells apparated not just in isolation, but some were in clonal clusters of 2–6 cells (Fig. 5J). These cells were still present at 162 h after the injury and were not observed in the resected tissue at 0 h (Fig. 5J and Supplementary Fig. 5E). The presence of this type of hepatocyte-like cell population during liver regeneration may suggest the existence of transdifferentiation processes from non-hepatocyte cells into hepatocytes. These cells resulted negative for the biliary epithelial cells and liver progenitor cells marker CK19, which was strictly expressed by bile duct cells in the portal areas (Supplementary Fig. 5F). Unexpectedly, we also observed some of these hepatocyte-like cells positive for both markers, CAG-EGFP and ≥1 Rainbow fluorophores (mOrange, mCherry or/and mCerulean) (Fig. 5J). Most of these marker-overlapping hepatocyte like-cells were multinucleated (Fig. 5J). Cell fusion, described in the liver22 could explain the origin of this type of unusual hepatocyte-like cells. In addition, given that some authors have documented that PHx conducted in rats increased multinucleated polyploid hepatocytes, we investigated changes in hepatocyte ploidy23. We found no alteration on liver ploidy after PHx, with 74% polyploid and 26% diploid hepatocytes within the liver lobule (Fig. 5K), similar values to those found in liver homeostasis (Fig. 1J). Like in control mice, Mid and PC zones had a similar majority of polyploid hepatocytes, 79% and 77% respectively, compared to PP areas, where 64% of the hepatocytes were polyploid (Fig. 5K). We did not observe elevation in the number of multinucleated polyploid hepatocytes, in fact, we found the opposite, with 95% of polyploid hepatocytes being mononucleated (Supplementary Fig. 5G). We confirm with these results our previous finding in the healthy and injured livers: that nuclear ploidy does not match hepatocyte ploidy (Supplementary Fig. 5G).

Hyperplasia and hypertrophy of midlobular hepatocytes repair the liver after coronavirus injury

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which caused the coronavirus disease 2019 (COVID-19) pandemic, targets mainly the lung due to the abundant expression of its receptor, angiotensin-converting enzyme 2 (ACE2), by respiratory epithelium. However, COVID-19 patients are also at risk of liver damage24,25. This finding is in line with studies that found high ACE2 receptor expression in the intrahepatic bile ducts26 and SARS-CoV-2 particles in the cytoplasm of hepatocytes from liver biopsies of COVID-19 individuals27. These data indicate that SARS-CoV-2 is capable of directly injuring the liver. The cytotoxicity of infection in the liver could lead to a systemic inflammatory response followed by multiorgan failure, which is observed in patients with severe COVID-1928. As SARS-CoV-2 is highly phylogenetically related to the murine native beta coronavirus Mouse Hepatitis Virus (MHV) Strain A5929, MHV-A59 has been used as a pre-clinical coronavirus model to reproduce many aspects of COVID-19 pathophysiology in mice30. Like SARS-COV-2, intranasal inoculation of MHV-A59 in mice produces acute respiratory distress syndrome, lymphopenia, multiorgan involvement, and systemic inflammation30. MHV-A59 also targets the liver via the CEACAM1 receptor, which, like the ACE2 receptor, is highly expressed in this tissue and, is also interferon-inducible31,32.

We used the MHV-A59 model to explore the pathological changes in the biliary cells and hepatocytes triggered by coronavirus infection and the regenerative capacity after this viral injury. We intranasally infected 10-week-old male Alb-CreERT2 Rosa26rbw mice with MHV-A59 after activating the Rainbow fluorophores with TMX. Mice sacrificed at 10 weeks of age were our control group and were considered week 0. At 1- and 4-weeks post-infection, we sacrificed the mice to assess liver architecture after viral damage and tissue repair respectively (Fig. 6A). We then used histological techniques to investigate hepatocyte clonal expansion, liver morphology, hyperplasia, hypertrophy and apoptosis (Fig. 6 and Supplementary Fig. 6). We used males in our study, as females were shown to be significantly more resistant to MHV-A59 and COVID-19 infection30,33,34. From days 1–7 after infection, the animals in the infected groups demonstrated respiratory illness symptoms, such as fever, shortness of breath, and a gradual decrease in body weight from days 0–7 compared with to PBS treated controls, which appeared normal over the entire period (Fig. 6B).

Fig. 6: Hyperplasia and hypertrophy of midlobular hepatocytes repair the liver after coronavirus injury.
figure 6

A Schematic showing the experimental design for coronavirus liver injury with MHV-A59. Created with BioRender.com. B Body weight curves in 10-week-old male mice during the first week of viral infection with MHV-A59 or PBS vehicle treatment; n = 3–4 mice per group. Data are presented as mean values +/− SEM. C Liver sections from mice after 1 and 4 weeks of MHV-A59 viral infection. Asteriks indicate immune cell infiltrates. CV central vein, PV portal vein. Scale bars, 100 μm. D Quantification of DAPI staining or total number of cells per field of view (10×) in liver sections from mice before and after 1 and 4 weeks of MHV-A59 infection. Data are presented as mean values +/− SEM. 5–8 photos analyzed per mouse; n = 3 mice per group. One-way ANOVA, Tukey’s multiple comparisons post-test. E Quantification of CAG-EGFP staining per field of view (10×) in liver sections from mice 1 and 4 weeks post-MHV-A59 infection. Data are presented as mean values +/− SEM. 5–8 photos analyzed per mouse; n = 3 mice per group. One-way ANOVA, Tukey’s multiple comparisons post-test. F Hepatocyte injury 1 week post-MHV-A59 infection in livers from mice. Left: Relative frequency in % of each clone size per area of the liver lobule and per Rainbow fluorophore from mice at 1 week post-MHV-A59 infection. Data are presented as mean values of the average PP, Mid and PC hepatocyte population +/− SEM. 6 liver lobule areas analyzed per mouse; 2 mice per time-point (4–5 photos per mouse). Middle: quantification of number of cells per clone (y-axis) and number of clones (x-axis) per area of the liver lobule and per Rainbow fluorophore from mice at 1 week post-MHV-A59 infection. Data are presented as mean values. 6 liver lobule areas analyzed from 2 mice per time point (4–5 photos per mouse). One-way ANOVA, Tukey’s multiple comparisons post-test. Right: Representative images of a liver lobule from mice at 1 week post-MHV-A59 infection; n = 2 mice. G Clonal expansion of hepatocytes 4 weeks post-MHV-A59 infection in livers from mice. Left: Relative frequency in % of each clone size per area of the liver lobule and per Rainbow fluorophore from mice at 4 weeks post-MHV-A59 infection. Data are presented as mean values of the average PP, Mid and PC hepatocyte population +/− SEM. 6 liver lobule areas analyzed per mouse; 2 mice per time-point (4–5 photos per mouse). Middle: quantification of number of cells per clone (y-axis) and number of clones (x-axis) per area of the liver lobule and per Rainbow fluorophore from mice at 4 weeks post-MHV-A59 infection. Data are presented as mean values. 6 liver lobule areas analyzed per mouse; 2 mice per time point (4–5 photos per mouse). One-way ANOVA, Tukey’s multiple comparisons post-test. Right: Representative images of a liver lobule from mice at 4 weeks post-MHV-A59 infection; n = 2 mice. Scale bars, 100 μm. H Quantification of proliferating Ki67 hepatocytes per field of view (10×) in each zone of the liver lobule from mice at week 4 post-MHV-A59 infection; Data are presented as mean values +/− SEM. 9–15 liver lobule areas analyzed from 2 mice per time point. One-way ANOVA, Tukey’s multiple comparisons post-test. I Hepatocyte hypertrophy after 4 weeks of MHV-A59 infection. Left: Hepatocyte size after tissue repair in liver sections from mice 4 weeks after MHV-A59 infection. Data are presented as mean values +/− SEM. Hepatocyte size (n = 53–91), from 3–4 liver lobule areas were analyzed from 2 mice. Unpaired Student’s t-test. Right: Hepatocyte size (n = 29) within the liver lobule from mice at 4 weeks after MHV-A59 infection. Data are presented as mean values +/− SEM. 3–4 liver lobule areas analyzed from 2 mice. One-way ANOVA, Tukey’s multiple comparisons post-test. (J) Representative CK19 IHC in liver sections from mice at 1 and 4 weeks after MHV-A59 infection. Arrows indicate cholangiocytes forming bile ducts around the portal triads; n = 2 mice. K Hepatocyte ploidy in the liver lobule of mice at 4 weeks after MHV-A59 infection. % of total polyploid and diploid hepatocytes within the liver lobule (left graph) and per area of the liver lobule (middle left graph). % of diploid hepatocytes (hepatocytes that express only 1 fluorophore) being mononucleated or multinucleated (middle right graph). % of polyploid hepatocytes (hepatocytes that express more than 1 fluorophore) being mononucleated or multinucleated (right graph). Data are presented as mean values +/− SEM. 5 photos analyzed per mouse; 2 mice per group. Source data are provided as a Source Data file.

We first examined liver changes by immunofluorescence after MHV-A59 infection (Fig. 6C). We identified a significant increase in total cells at 1- and 4-weeks post-infection compared to control mice by quantifying the number of DAPI stained nuclei (Fig. 6D). In addition, we detected more total cells at 1 week than 4 weeks post-infection (Fig. 6D). Similarly, we found higher CAG-EGFP+ area (%) at 1 week than 4 weeks after infection, indicating the generation and elimination of non-hepatocyte populations respectively (Fig. 6E). We also observed hepatocyte clonal expansion at 4 weeks post-infection compared to 1 week, indicative of liver repair processes (Fig. 6F, G). Particularly, hepatocyte clones within the liver lobule increased in number and size (Fig. 6F, G). Larger clones of 4 to 7 cells were detected at 4 weeks post-infection and their relative frequency was higher than at 1-week post-infection (Fig. 6F, G). These larger clonal clusters were in the Mid areas, followed by PC and PP zones (Fig. 6F, G). These findings were consistent with our observations in other liver injury models, such as, CCl4 damage and PHx. In line with this observation, we also observed more proliferating Ki67 hepatocytes in the Mid areas at week 4 post-infection (Fig. 6H). Liver regeneration after 4 weeks was also accompanied by hepatocyte hypertrophy, with Mid hepatocytes being the largest (Fig. 6I). On week 1 post-infection, H&E staining and TUNNEL assay demonstrated massive hepatic necrosis and apoptosis respectively (Supplementary Fig. 6A–B). Immunohistochemical analysis of CD68 and CD45 markers detected considerable inflammatory infiltrates mainly distributed within the liver parenchyma, from central veins to Mid areas of the liver lobule (Supplementary Fig. 6D–E). Although, we also found CD68 and CD45 expressed by non-parenchymal cells, in the hepatic sinusoids, suggesting activation of Kupffer cell and immune response. Immunohistochemical staining of CK19 showed no signs of cholestasis nor ductular reaction originated by cholangiocytes (Fig. 6J). Ki67 staining of consecutive liver slides showed that cell proliferation was restricted to immune cells (Supplementary Fig. 6C). On week 4 post-infection, liver lobule architecture did not show any significant histological alteration with respect to the control group indicating that the mouse adult liver preserves its regenerative capacity after coronavirus infection (Fig. 6C and Supplementary Fig. 6). Finally, we quantified the ploidy status of hepatocytes in each zone of the liver lobule after 4-weeks of MHV-A59 infection to evaluate whether hepatocytes developed during the recovery phase underwent ploidy changes (Fig. 6K). We found an increase in diploid hepatocytes with respect to liver homeostasis (Fig. 1J), with 39% being diploid vs. 61% polyploid (Fig. 6K). Like in control mice, Mid and PC zones had a majority of polyploid hepatocytes, with a 65% and 72% respectively compared to PP areas, where 42% of the hepatocytes were polyploid. Contrastingly, PP were mostly diploid (58%), followed by Mid (35%) and PC (28%) (Fig. 6K). Most hepatocytes were mononucleated regardless of ploidy status, with 81% and 89% of polyploid hepatocytes being mononucleated (Fig. 6K). So, with this model of coronavirus damage, we confirm the previous findings in healthy and injured livers: that nuclear ploidy does not match hepatocyte ploidy (Fig. 6K).

Cell–cell communication analysis reveals that Gal-9-CD44 pathway may play a critical role in regulating liver homeostasis after injury by attenuating immune cell infiltration and stimulating hepatocyte proliferation

To highlight the underlying mechanisms mediated by immune cells, endothelial cells and cholangiocytes in orchestrating the reparative response together with hepatocytes to drive liver growth after injury, we performed an analysis of cell–cell communication from our scRNA-seq dataset after acute CCl4 treatment (Fig. 7 and Supplementary Figs. 1012). We analyzed liver preparations from Alb-CreERT2 Rosa26rbw mice on day 6 after CCl4 insult, when we observed peak proliferation of hepatocytes (Supplementary Fig. 7D). As expected, our cell–cell communication analysis revealed increased total cell–cell communication in CCl4-treated mice, however, the hepatocyte communication with other cells was surprisingly reduced by injury (Fig. 7B and Supplementary Fig. 11A).

Fig. 7: Changes in Intracellular communications network in the liver niche during CCl4 treatment.
figure 7

A UMAP clustering of single cell data with Seurat from liver preparations isolated from Alb-CreERT2 Rosa 26rbw mice after 6 days of being treated with a single dose of vehicle (control) or CCl4 (CCl4). Liver cells were clustered in two dimensions using the UMAP dimensionality reduction technique and annotated by cell type. UMAP representation of 11 distinctive cell clusters from liver cells. Top 3 differentially expressed marker genes are shown for each cell cluster. N = 2 Alb-CreERT2 Rosa26rbw mice per group. B Circle plot highlighting the differential number of ligand-receptor (L-R) interactions and interaction strength between control and CCl4 treated animals. N = 2 Alb-CreERT2 Rosa26rbw mice per group. C Chord diagram showing upregulated signaling ligand-receptor pairs in CCl4 condition. Each link indicates a ligand-receptor pair. The root of each arrow is the ligand-expressing cell type, and the tip of each arrow is the receiving cell. N = 2 Alb-CreERT2 Rosa26rbw mice per group.

Interestingly, we identified cholangiocytes as the liver cell type in CCl4-injured mice that interacted most with themselves and other hepatic cells through expression of ligands-receptors (Fig. 7B). Particularly, cholangiocytes showed strong interactions with other cholangiocytes, followed by monocytes and endothelial cells. Monocytes also presented a notable interaction with themselves, dendritic cells and endothelial cells. Dendritic cells and endothelial cells also displayed increased interactions with other cells. By contrast, CCl4 significantly reduced ligand-receptor interactions between hepatocytes and immune cells, including monocytes and T-cells, as well as endothelial cells (Fig. 7B).

Importantly, our analysis of ligand-receptor pairs found that Lgals9, the gene that encodes the protein Gal-9 which is highly expressed in the liver35, was the most upregulated gene by CCl4 in many cell types, including hepatocytes, cholangiocytes, dendritic cells, endothelial cells, macrophages, monocytes and T-cells (Fig. 7C). Several receptors or surface binding partners have been reported for Gal-9 including the adhesion molecule cluster of differentiation 44 (CD44)35, which is also upregulated in our cell–cell interaction analysis in the same hepatic cell types where Lgals9 was overexpressed except in hepatocytes (Fig. 7C). These results elucidate Gal-9 as a major mediator of communication between cell types to coordinate the reparative response to liver injury.

We also investigated others enriched pathways in livers from CCl4 and control mice as shown Supplementary Fig. 11B to assess the cellular and molecular mechanism that regulate liver regeneration in response to injury. We found that numerous signaling pathways were upregulated by different cell populations 6 days after acute CCl4 damage (Supplementary Fig. 11B). Most of these pathways belong to signaling cascades associated to regulate the inflammatory and immune responses (CD2236, MHC-I37, CD5238, CHEMERIN39, CD22640, NECTIN41, MHC-II42, SN, CD2343, CD8044, TNF45, CSF46, APRIL47, IL1048, PVR49, AGT50, IL1651, CEACAM52, CD20053, OCLN54, ACTIVIN55, ANNEXIN56, SELPLG57, CD8644, CXCL58, SEMA459, PECAM160, TGFb61, BST262, SELL63, LAIR164, ITGAL-ITGB265, JAM66, CCL67, MIF68, VCAM69, ICAM70, COMPLEMENT71, PROS72 and CD4573) (Supplementary Fig. 11B). Notably, many of the upregulated signaling pathways are known to be also activated during liver repair and regeneration (CSF46, ACTIVIN55, ANNEXIN56, VCAM69, ICAM70, TNF45, TGFb61 and CEACAM52). Moreover, other pathways that are also drivers of hepatic remodeling resulted upregulated after CCl4 including WNT74, HGF75, TWEAK76, GRN77, NOTCH78, IGF79, COLLAGEN80 and LAMININ81 (Supplementary Fig. 11B). The upregulation of several known liver repair pathways confirms that our data set robustly assessed the response of liver to injury.

Taken together, our results indicate that CCl4 causes liver damage through free radical mediated inflammatory processes that trigger the death of PC hepatocytes initially at day 2 post CCl4, as we observed in Fig. 2C, G. This CCl4 toxicity causes macrophages and monocytes infiltration which facilitate the removal of cellular debris generated by necrosis and apoptosis of PC hepatocytes (Fig. 2H). At day 3 post CCl4, PC hepatocytes and PC clones originate by division of undamaged Mid hepatocytes, which reached a peak of hepatocyte proliferation at day 6 (Supplementary Fig. 7C, D). At this time point, cholangiocytes, monocytes and endothelial cells became very active, increasing significantly their interaction through the upregulation of numerous outgoing (ligands) and incoming (receptors) signaling patterns, as shown in greater detail in Fig. 7 and Supplementary Fig. 11C.



Source link

Related Articles

Leave a Reply

[td_block_social_counter facebook="beingmedicos1" twitter="being_medicos" youtube="beingmedicosgroup" style="style8 td-social-boxed td-social-font-icons" tdc_css="eyJhbGwiOnsibWFyZ2luLWJvdHRvbSI6IjM4IiwiZGlzcGxheSI6IiJ9LCJwb3J0cmFpdCI6eyJtYXJnaW4tYm90dG9tIjoiMzAiLCJkaXNwbGF5IjoiIn0sInBvcnRyYWl0X21heF93aWR0aCI6MTAxOCwicG9ydHJhaXRfbWluX3dpZHRoIjo3Njh9" custom_title="Stay Connected" block_template_id="td_block_template_8" f_header_font_family="712" f_header_font_transform="uppercase" f_header_font_weight="500" f_header_font_size="17" border_color="#dd3333"]
- Advertisement -spot_img

Latest Articles