Fibrosis-resistant/-susceptible inbred mouse strain model
First, we characterized fibrosis susceptibility in three common inbred mouse strains (FVB, C57Bl/6, and BALB/c) subjected to chronic administration of hepatotoxin thioacetamide (TAA) for 6 weeks. FVB demonstrated resistance; BALB/c strain developed the most significant fibrosis; and C57Bl/6 mice showed intermediate susceptibility to liver fibrosis (Supplementary Fig. 1).
Chronic hepatotoxin-induced fibrosis models are characterized by repetitive liver injury, which leads to gradual fibrosis progression over an extended period of time. We hypothesized that such dramatically divergent strain phenotypes might be determined by strain-specific differences in response to each acute insult. To test this, we established a model of recovery from a single sub-lethal dose of TAA, and compared the FVB, C57Bl/6, and BALB/c strains’ responses 1, 3, 5, and 8 days thereafter (Fig. 1a). Biochemical quantification of hepatic collagen revealed readily detectable progressive collagen deposition into BALB/c livers starting from day 3 and reaching maximum (142% above control group levels, p < 0.001) on day 8 post TAA. In contrast, both FVB and C57Bl/6 mice developed significantly less fibrosis, with maximum collagen levels reaching 77% and 88%, respectively, above their healthy controls (Fig. 1b). Upon histological examination, connective tissue deposits were readily detected by Sirius Red staining starting on day 8 in pericentral areas of BALB/c mice, but not of FVB or C57Bl/6 mice (Fig. 1d). The degree of hepatotoxin-induced injury was similar between all three strains studied, as measured by cumulative serum alanine aminotransferase (ALT) levels 12–96 h post TAA (“area under curve” analysis, Fig. 1c). Overall, based on data in this single-dose TAA injury model, we were able to recapitulate strain differences in susceptibility to fibrosis observed in the chronic TAA injury model (BALB/c > C57Bl6 > FVB, Supplementary Fig. 1). Importantly, similar results were obtained in FVB and BALB/c mice after acute injury with another hepatotoxin, carbon tetrachloride (CCl4), suggesting that these strain differences were not toxin specific (Supplementary Fig. 2).
Delayed dead hepatocyte clearance precedes fibrotic response
Further histological examination revealed that by 24 h, TAA caused a similar degree of pericentral hepatocyte cell death in ~50% of hepatocytes in each strain. However, striking differences were observed in the speed with which the necrotic masses were cleared. In the resistant FVB strain, necrotic hepatocytes were rapidly cleared and disappeared completely by day 3, while in BALB/c they persisted even after 5 days (Fig. 2a). C57Bl/6 strain demonstrated necrotic masses clearance rate similar to fibrosis-resistant FVB with a delay of about 24 h. All subsequent experiments were performed in FVB and BALB/c mice, which demonstrated the most divergent responses. Macroscopically, at day 8 of recovery from TAA injury, BALB/c livers demonstrated increased stiffness and patchy whitish lesions throughout the liver parenchyma, while FVB livers appeared healthy and virtually indistinguishable from untreated controls (Fig. 2b). In both strains, recovery from TAA injury was accompanied by a robust up-regulation of pro-fibrogenic messenger RNA (mRNA) encoding for TGFβ1, pro-collagen α1(I), TIMP-1, and matrix metalloproteinase-2 (MMP-2), with the peak at day 3, except for TIMP-1, which was induced after 1 day (Fig. 2c). Surprisingly, the degree of transcript up-regulation was nearly identical between the two strains in the early phase (0–3 days), but differed significantly only at the late time-points (5 and 8 days). After day 5, most pro-fibrogenic mRNAs normalized in FVB mice, but remained elevated in BALB/c, with the strain differences most apparent at day 8 post injury (Fig. 2c).
Impaired efferocytosis function in fibrosis-susceptible mice
The major cells responsible for dead cell recognition and clearance in vivo are macrophages (“professional phagocytes”), which prompted us to investigate temporo-spatial changes in macrophage infiltrate. Both strains demonstrated similar immune cell recruitment in early injury (12–24 h post-TAA) stages, as assessed via hepatic mRNA expression of pan-leukocyte marker CD45 and key macrophage/monocyte markers F4/80, Clec4F, CD68, and CD11b/c (Supplementary Fig. 3a). However, notable temporo-spatial differences became apparent in the intermediate (3 days) and late (8 days) repair phase. In situ F4/80 staining showed that MΦ infiltrated necrotic areas at day 3 in FVB mice, and these infiltrates cleared completely by day 8. In contrast, in BALB/c mice macrophages accumulated at the edge of necrotic areas at day 3, and remained in large numbers even at day 8, suggesting incomplete efferocytosis (Supplementary Fig. 3b). Immune cell infiltrates in both strains also stained positive for myeloid cell markers Gr-1 (monocytes and granulocytes) and CD11b (monocytes) (Supplementary Fig. 3c, d). Double immunostaining for F4/80 and desmin revealed that active fibrogenesis likely occurs within these persistent macrophage infiltrates in BALB/c mice, as evidenced by marked accumulation of desmin+ HSCs/MFs; in contrast, desmin+ HSCs/MFs retained normal peri-sinusoidal distribution in resistant FVB strain (Fig. 3a).
In order to directly assess whether function of dead cell removal by phagocytes (efferocytosis) is indeed impaired in the livers of the fibrosis-susceptible strain, we performed an in vivo phagocytosis assay via administration of strain-matched fluorescently labeled apoptotic thymocytes into FVB and BALB/c mice (both healthy and after TAA-induced liver injury). We then measured their engulfment (localization within macrophages) 1 h later. Analysis of apoptotic cell distribution in relation to macrophage markers revealed that liver homing of apoptotic cells, as well as their engulfment rate by F480+-resident macrophages, was significantly lower in injured livers of BALB/c mice compared to FVB (Fig. 3b, c). Apoptotic cell numbers were similar in livers of healthy FVB and BALB/c, and demonstrated similarly high co-localization with F4/80+ macrophages (>85% engulfed). After liver injury, apoptotic cell homing and percent of engulfed apoptotic cells were depressed in both strains; however, the magnitude of change was remarkably more pronounced in BALB/c compared to FVB. Thus, total apoptotic cell counts decreased by 57% (BALB/c) versus 31% (FVB), whereas percent of apoptotic cell engulfed by macrophages decreased by 43.5% (BALB/c) versus 24.7% (FVB) compared to healthy strain-matched controls (Supplementary Table 3). Proportion of apoptotic cells co-localized with infiltrating Gr-1+ myeloid cells in injured livers was very low (>5%) compared to that of F4/80+ macrophages, and did not differ between strains (Supplementary Table 3).
Finally, to identify phagocytic genes potentially responsible for inefficient efferocytosis in the fibrosis-susceptible BALB/c strain, we profiled hepatic expression of phagocytosis-related genes using mini-array. Among the 84 relevant genes analyzed, only four genes were differentially expressed between strains at baseline (Fig. 3d). In contrast, 48 h after injury, 20 genes were upregulated in response to injury in FVB mice compared to BALB/c (Fig. 3e, p < 0.05, see Supplementary Tables 4 and 5 for a complete list of genes). Upon additional cross-validation using TaqMan PCR for the top 10 differentially regulated genes, five genes were confirmed as significantly different between strains post injury (Supplementary Table 6). Among these were genes encoding phagocytic receptors (Cd14, Marco); recognition and engulfment molecules (Csf1); and phagosome maturation (Serpine1, Tgm2). Thus, delayed clearance of necrotic masses (Fig. 2a), inefficient efferocytosis (Fig. 3b, c), and persistence of macrophage infiltrates (Supplementary Fig. 3) in the injured livers of fibrosis-susceptible BALB/c mice associated with failure to efficiently upregulate multiple genes involved in the phagocytic pathway.
Phagocyte depletion abrogates resistance to fibrosis
To clarify the mechanistic role of phagocytic liver macrophages and infiltrating myeloid cells in fibrosis susceptibility, we used both injection of clodronate-loaded liposomes (CLO), which efficiently eliminates tissue macrophages with high phagocytic activity (Supplementary Fig. 4a), and antibody-mediated depletion of myeloid cell subsets. First, we depleted the macrophages either 1 day before or 1 day after TAA-induced injury by a single intraperitoneal (i.p.) injection of clodronate-containing liposomes, to eliminate (1) resident phagocytic macrophages or (2) resident and recruited/infiltrating phagocytic macrophages, respectively (Fig. 4a) and compared fibrotic responses at the late (8 days post injury) time-point. Based on quantitative analysis of collagen deposition in these animals, macrophage depletion in resistant FVB mice rendered them completely susceptible to fibrosis, and caused only minor changes in collagen deposition in BALB/c mice (Fig. 4b). Histologic examination revealed that MΦ-depleted FVB mice failed to clear necrotic masses of hepatocytes, which persisted at day 8 post injury (Supplementary Fig. 4b) and developed fibrotic lesions histologically similar to that observed in fibrosis-susceptible BALB/c strain, regardless whether depletion was performed before or after injury (Fig. 4c). This was accompanied by a dramatic increase in pro-fibrogenic gene expression: TGFβ1, pro-collagen α1(I), and TIMP-1 were upregulated 2–6-folds above controls in FVB mice and did not normalize by day 8 in both strains with MΦ depletion (Fig. 4d). Upon depletion of phagocytes, there was virtually no difference in fibrotic response between these two mouse strains upon assessment by histology or quantitative fibrosis measures (Fig. 4). Importantly, we were able to reproduce these results when chronic MΦ depletion was performed concurrently with chronic TAA administration for 6 weeks. Resistance to fibrosis was compromised in FVB mice receiving weekly CLO injections, which developed significant hepatic fibrosis, as evidenced by increase in collagen deposition, histological signs of bridging fibrosis, and increase in HSC activation marker α-SMA, compared to controls treated with TAA and vehicle [phosphate-buffered saline (PBS)]-loaded liposomes (Supplementary Fig. 5). Since both resident macrophage and infiltrating Gr-1(+)/CD11b(+) myeloid cells appeared to accumulate within post-necrotic lesions (Fig. 3, Supplementary Fig. 3), we also performed myeloid cell subset depletions using lineage-specific antibody during recovery of resistant FVB mice from TAA injury (Fig. 4a). Cell depletion using Gr-1 (but not CD11b or Ly-6G)-specific antibody resulted in persistence of dead hepatocyte at day 8 post injury, and compromised fibrosis resistance in FVB mice (Fig. 4e). Of note, administration of Gr-1 (but not CD11b or Ly-6G)-specific mAB led to HSC activation on day 8 post TAA (Supplementary Fig. 6a) and increased hepatic collagen deposition as determined biochemically via hydroxyproline (Supplementary Fig. 6b). These data collectively suggest that both resident liver macrophages [Kupffer cells (KCs)] and infiltrating Gr-1(+) myeloid cell subset are functionally required for efficient clearance of dead hepatocytes, and appear to protect from ensuing fibrotic responses.
Hepatocyte-derived mito-DAMPs compromise fibrosis resistance
The results described above suggested to us that prolonged exposure to the certain DAMPs released from persisting dead hepatocytes in fibrosis-susceptible strain (or when MΦ function is compromised in resistant FVB mice) may be directly responsible for triggering exaggerated fibrotic responses. One such group of molecules leaking from damaged mitochondria (collectively called mito-DAMPs) is particularly immunogenic and has been implicated in various diseases8. Indeed, we detected transient 3-fold elevation in serum levels of mtDNA (a marker of mito-DAMPs release) 2 days post-TAA injury in fibrosis-susceptible BALB/c mice, but not in resistant FVB mice (Fig. 5a). Importantly, this was not due to strain differences in mtDNA genome copy number, which was similar in FVB and BALB/c livers relative to nuclear genome (Supplementary Fig. 7d, e). To test if mito-DAMPs were directly pro-fibrogenic, we purified liver mitochondria, prepared mito-DAMPs from them, and injected them into resistant FVB mice on the second and fourth day post-TAA injury to mimic prolonged exposure to mito-DAMPs due to inefficient efferocytosis of damaged hepatocyte in vivo (Fig. 5b). The mito-DAMP dose chosen contained 9.5 μg of mtDNA, which corresponds to 10% of the whole liver and in all likelihood reflects pathophysiologic levels based on 40–50% hepatocyte death observed in response to TAA in our system and kinetics of systemic exogenous mito-DAMPs exposure in vivo (Supplementary Fig. 7c). While control FVB mice, as expected, were able to repair the liver injury in a scarless manner by day 8 post injury, FVB mice injected with mito-DAMPs developed significant liver fibrosis, with readily detectable collagen deposits, morphologically characterized as incomplete fibrotic septa (Fig. 5c). Quantitatively, the overall increase in collagen deposition in mice injected with mito-DAMPs was significant, with a 2.5-fold increase as determined via connective tissue square morphometry, and a 27.45% increase as determined via biochemical hydroxyproline assay, respectively (p < 0.05, Fig. 5d, e). Areas of collagen deposition in livers of mito-DAMP-treated mice overlapped with a massive increase in immunopositivity for α-SMA, an HSC activation marker. Moreover, pro-fibrogenic gene transcript levels of TGFβ1, pro-collagen α1(I), and TIMP-1 in livers of mito-DAMP-treated mice remained elevated 3–4-folds at day 8 post injury (Fig. 5f). In order to validate these findings in a second, mechanistically different model of steatohepatitis, we injected mice primed with short-term methionine- and choline-deficient diet (MCD) feeding (2 weeks) with a single dose of purified mito-DAMPs (9.5 μg/mouse, i.p.) and examined the livers 24 h later; in situ immunostaining for α-SMA revealed massive activation of peri-sinusoidal HSCs throughout the liver lobule, with a 2.6-fold increase in α-SMA area morphometrically (Fig. 5g).
Mito-DAMPs directly trigger pro-fibrogenic HSC activation
In order to investigate whether pro-fibrotic action of mito-DAMPs may be due to direct effect on HSCs, a main fibrogenic effector cell in the liver, we freshly isolated HSCs and incubated them with increasing concentrations of mito-DAMPs (0.25, 0.5, and 0.75 μg/ml) in vitro. Twenty four hours after the addition of mito-DAMPs, HSCs demonstrated dose-dependent changes in morphology characteristic of their activation, with the loss of lipid droplets, assumption of an MF-like appearance, and robust increases in immunopositivity for α-SMA (Fig. 6a). HSC proliferation, as assessed by MTT [3-(4,5-dimethylthiazol-2-yl)- 2,5-diphenyltetrazoliumbromide] assay, was dose-dependently increased in the presence of exogenously added mito-DAMPs (Fig. 6b). Concomitantly, primary HSCs upregulated pro-fibrogenic gene expression of pro-collagen α1(I), TGFβ1, and TIMP-1 (2–4-folds over controls, Fig. 6c). Pre-incubation of mito-DAMPs with DNase I immediately prior to the addition to HSC cultures efficiently depleted mtDNA and significantly attenuated its pro-fibrogenic activity by up to 50% compared to intact mito-DAMP preparations, as assessed via cell proliferation, α-SMA induction, and fibrogenic gene expression (Fig. 6d–h). Interestingly, primary resident liver macrophages (Kupffer) cells failed to respond to equivalent doses of mito-DAMPs in vitro, both at baseline and upon activation by lipopolysaccharide (LPS) or interferon-γ (IFNγ), as measured by secretion of tumor necrosis factor-α (TNF-α), interleukin-1β (IL-1β), or nitric oxide (NO) (Supplementary Fig. 8). Taken together, these results suggest that exposure to exogenous mito-DAMPs directly triggers fibrogenic activation in HSCs, with mtDNA as a major active component, as evidenced by characteristic changes in morphology, up-regulation of α-SMA, increased cell proliferation, and pro-fibrogenic transcriptional activity.
Circulating mito-DAMPs are elevated in human NAFLD/NASH
In order to test if circulating mito-DAMPs levels are increased in human disease, we performed a survey of mtDNA levels in the serum of 27 patients (cohort I) with biopsy-confirmed non-alcoholic fatty liver disease (NAFLD) via PCR amplification of a short fragment in the D-loop region of mtDNA. The majority of patients (20/27) in this cohort had active NASH, defined histopathologically as NAFLD activity score (NAS) ≥ 4, and various degrees of fibrosis ranging from F0 (no fibrosis) to F4 (cirrhosis). Twelve healthy controls of different ages (24–63 years old) were analyzed for comparison (Table 1). Serum mtDNA levels were increased 7-fold on average in NAFLD patients with different stages of disease when compared to healthy subjects (p < 0.0001), with mtDNA levels reaching up to a 17-fold increase above average normal values (Fig. 7a). Circulating mtDNA values in healthy subjects did not appear to vary with age or sex. Association with histological parameters of disease progression (NAS score and fibrosis stage on liver biopsy) was tested in an additional 114 patients with NAFLD/NASH (cohort II). Circulating mtDNA levels were further increased in patients with active NASH (NAS score 4–8, versus minimal disease activity NAS score 0–3, p = 0.0334, Fig. 7b) and particularly in those with the significant histological signs of fibrosis [F2–4, 482.4 ± 62.66, 95% confidence interval (CI), 356.2–608.6] on biopsy, compared to patients with minimal/no fibrosis (F0–1, 253.4 ± 26.45, 95% CI, 200.6–306.2, p = 0.0003, Fig. 7c).