Scar-Associated Macrophages Are a Major Source of Hepatic Matrix Metalloproteinase-13 and Facilitate the Resolution of Murine Hepatic Fibrosis¹

We have recently demonstrated an important role for macrophages in the normal resolution of hepatic fibrosis using a model of CCl₄-induced liver injury followed by spontaneous recovery in rodents (1). Using a conditional ablation mouse model to selectively deplete macrophages during the resolution phase of hepatic fibrosis, we observed failure to remodel interstitial collagen, particularly perisinusoidal fibrosis, as a result of the selective depletion of scar-associated macrophages (SAMs)³ (1). We were therefore interested to discover the key effector molecules and mechanisms by which macrophages mediate this process.

Macrophages have been shown to produce gelatinases (matrix metalloproteinase (MMP)2, MMP9), metalloelastase (MMP12), matrilysin (MMP7), and collagenases (MMP1 and MMP13) in different circumstances (2, 3), allowing them to degrade complex extracellular matrices. Further support for an important role of the macrophage in matrix remodeling comes from in situ hybridization studies showing abundant MMP mRNA within macrophages in the rheumatoid synovial pannus (4), in atherosclerotic plaques (5), and in the lung during injury (6).

Remodeling of fibrillar collagen in rodents has been widely attributed to the action of MMP13, the main collagenase identified in this order. This consensus is based on unproductive attempts to detect the presence in rodent tissue of MMP1 (the major interstitial collagenase in humans) which suggests that MMP1 may be functionally substituted in rodents by MMP13 (7). Identifying the key collagenase(s) that mediates the regression of rodent liver fibrosis is pivotal to understanding models that are used in proof of concept studies of antifibrotic compounds in vivo (8, 9, 10).

MMP13 has potent proteinase activity against a wide variety of extracellular matrix components, particularly fibrillar collagens and gelatins. MMP13 has a central position in the MMP activation cascade, both activating and being activated by several MMPs (7, 11, 12). Active MMP13 is inhibited in a 1:1 stochiometric fashion by tissue inhibitors of metalloproteinases (TIMPs) 1, 2, and 3). Further regulation of MMP13 synthesis is mediated by cytokines (e.g., TGF-β), growth factors and probably also by interaction with the extracellular matrix via integrin-mediated pathways, thereby providing the matrix with a mechanism by which it can regulate its own composition.

In progressive liver disease in human and animal models of liver fibrosis, the level of expression of MMP1/MMP13 in most cases does not appear to alter significantly, whereas the expression of TIMP-1 and -2 is markedly increased (13, 14). In contrast, enhanced mmp13 mRNA transcription during resolution of CCl₄-induced rat liver fibrosis has been reported (15). In this 8-wk model, mmp13 expression was observed early and transiently between days 5 and 7 after cessation of injury.

The cellular origin of MMP13 remains controversial. Contrasting in vivo models have suggested that several cell types have the capacity to produce MMP13 in certain circumstances, including hepatic stellate cells (HSCs; Ref. 15), Kupffer cells (16), and stem cells (17) but with no clear consensus. Early liver injury, in particular, is characterized by a rich inflammatory infiltrate which includes recruited monocyte-macrophages. These may represent an alternative source of MMP13 or other enzymes with interstitial collagenase activity. Indeed, the initial surge in MMP13 expression in a model of acute liver toxicity appeared to parallel TNF-α expression and the presence of tissue inflammation (18).

We hypothesized that SAMs might represent the elusive cellular source of collagenolytic MMP13 in hepatic fibrosis. In this study, we show a temporal correlation between the level of MMP13 expression and the presence of macrophages in the injured and recovering liver after fibrotic injury and further localize MMP13 in vivo to SAMs. In addition, we have previously shown that conditional macrophage ablation inhibits remodeling of hepatic fibrosis (1), and in this study we demonstrate a similar fact when MMP13 is deleted. This supports a critical role for macrophage-derived MMP13 in mediating the regression of hepatic fibrosis.

Animals were housed in standard sterile conditions with free access to chow and water. All manipulations and procedures were undertaken in accordance with United Kingdom Home Office license regulations and protocols approved by the Standing Committee on Animals of Harvard Medical School.

Liver injury was induced as previously described (19). In brief, cohorts of 12 Sprague–Dawley rats were injected i.p. with 0.2 ml/100 g sterile CCl₄ dissolved in a 1:1 ratio with olive oil twice weekly for 4, 6, and 8 wk to generate an early reversible fibrosis and early and established cirrhosis, respectively. In addition, a cohort was treated for 12 wk to establish advanced micronodular cirrhosis which undergoes only partial resolution during 1 year of follow-up. For each model, animals were euthanized, and livers were harvested at peak fibrosis (immediately after the final injection of CCl₄) and at 3, 14, 28, 84, 168, and (in the 12-wk model) 366 days of spontaneous recovery (n = 4 at each time point in each model). Three normal, untreated rat livers were also harvested for use as controls in individual experiments. Harvested livers were split and fixed in formalin for subsequent immunohistochemical studies or snap frozen in liquid nitrogen for biochemical and molecular analysis.

Details of the generation of the CD11b-DTR mice have been described previously (1). In brief, adult CD11b-DTR mice (FVB/N) were injected with 0.25 μl/g CCl₄ i.p. twice weekly for 12 wk. During the 12th week, the mice were given by tail vein injection either: 1) diphtheria toxin (DT; 10 ng/g) immediately after the first injection of CCl₄ that week, DT 24 h later, and DT an additional 48 h later to coincide with the next injection of CCl₄; 2) DT (25 ng/g i.p.) over a similar schedule; or 3) 100 μl of PBS (i.v.) as control to coincide with the DT injections. Livers were harvested 24 h after the final dose of DT or PBS.

MMP13-deficient mice (mmp13^−/−) were a gift from Dr. S. Krane (Harvard Medical School, Boston, MA). The mmp13^−/− mice were generated by gene targeting in embryonic stem cells as previously described (20). Genotype of the mice was verified by a two-step PCR using Taq polymerase (Qiagen) and the following primer pairs: primer 1/primer 3; primer 2/primer 3. Sequences were as follows: primer 1 (phosphoglycerate kinase) 5′-CGAAGGAGCAAAGCTGCTA-3′; primer 2 (exon 5F) 5′-TTTATTGTTGCTGCCCATGAG-3′; primer 3 (exon 6R) 5′-AGTTTCTCCTCGGAGACTGGT-3′. Conditions for PCR were 94°C for 3 min; then 40 cycles of 94°C for 45 s, 50°C for 45 s and 72°C for 2 min; then finally 72°C for 10 min.

Liver fibrosis was induced in cohorts of sex- and age-matched mmp13^−/− and wild-type (WT) C57BL/6 mice by 4 wk, three-times-weekly i.p. administration of escalating doses of CCl₄ dissolved in sterile olive oil as follows: week 1, 0.125 ml/kg; week 2, 0.25 ml/kg; week 3, 0.5 ml/kg; week 4, 1 ml/kg. This escalating 4-wk CCl₄ protocol results in a rapid induction of advanced liver fibrosis (Q. Anstee, Imperial College, London, U.K., unpublished observations). Animals were euthanized at peak fibrosis (24 h after the final dose of CCl₄) and after 1, 3, and 5 days of spontaneous resolution (n = 5 at each time point). Four mmp13^−/− and WT mice received olive oil only and served as vehicle controls. Harvested livers were split and fixed in formalin for subsequent immunohistochemical studies or snap frozen into liquid nitrogen for biochemical and molecular analysis.

RNA was purified from CCl₄-treated and olive oil-treated (control) livers using the RNAeasy method (Qiagen). First-strand cDNA synthesis was undertaken using random primers and murine Moloney leukemia virus reverse transcriptase (Promega). Primers and FAM-TAMRA probes were designed using the TaqMan Primer Express program, and real-time quantitation was obtained using the PE Applied Biosystems 7700 Sequence Detection System. The following were primers/probes for rat: mmp13, sense 5′-GGTTGAGCCTGAACTGTTTTTGA-3′, antisense 5′-CTCGTATGCAGCATCCACATG-3′ and probe 5′-AGTCCTTTTGGCCAGAACTTCCC-3′ (Oswel Research Products); GAPDH, sense 5′-GGCCTACATGGCCTCCAA-3′, antisense 5′-TCTCTCTTGCTCTCAGTATCCTTGC-3′, and probe 5′-AGAAACCCTGGACCACCCAGCCC-3′ (Oswel Research Products). The following were primers/probes for mouse: mmp2, sense 5′-CCCCATGAAGCCTTGTTTACC-3′, antisense 5′-TTGTAGGAGGTGCCCTGGAA-3′, and probe 5′-GAATGCTGATGGACAGCCCTGCA0-3′ (Sigma-Genosys). Primers and probes for mouse MMP8 and MMP14 were custom designed and supplied by Primer Design. 18S rRNA Control Reagents (VIC probe) were manufactured by Applied Biosystems.

First-strand cDNA (1 μl = 10 ng of RNA), 0.3 μM concentrations of primers and 0.3 μM concentrations of probe were used in a 25-μl real-time TaqMan reaction using 2× Universal Master Mix (Applied Biosystems). Conditions were as follows: 50°C for 2 min; then 40 cycles of 95°C for 10 min, 95°C for 15 s, and 60°C for 1 min. All reactions were performed in triplicate. After detection of the threshold cycle for each mRNA in each sample, relative concentrations were calculated and normalized to control genes GAPDH or 18S. Products were run on an agarose gel to confirm a single band of the correct size.

mmp13 and timp1 transcripts were detected in situ using oligonucleotide probes purchased from GeneDetect. All other reagents were supplied by DakoCytomation for oligoprobes and Roche for riboprobes unless stated otherwise.

mmp13 mouse antisense GreenStar digoxigenin (dig)-labeled oligoprobe hybridizes to nt 158–205 located within the coding sequence of NM_008607. Two antisense riboprobes and their sense counterparts were also synthesized by cloning nucleotide sequences from positions 24–440 and 805–1240 into the pCR^R2.1 linearized plasmid vector. After sequencing, dig-labeled probes were generated using the T7 promoter system (Promega). Both the size and extent of dig-labeling of probes were confirmed.

timp1 mouse antisense GreenStar dig-labeled oligoprobe hybridizes to nt 294–341 located within the coding sequence of NM_011593.

Tissue sections of 7 μm were cut under RNase-free conditions, dewaxed in xylene, and rehydrated through graded alcohols, then permeabilized with proteinase K at room temperature for 5 min. The protease was quenched in glycine (0.2% w/v in 1× PBS); then sections were rinsed in PBS and postfixed in paraformaldehyde (4% w/v in PBS), and endogenous peroxidase was blocked with hydrogen peroxide (1% w/v in methanol). Sections were prehybridized with mRNA in situ hybridization solution for 2 h at 37°C. Probes (sense control and antisense) were added at 200 ng/ml in hybridization solution overnight at 37°C. After stringency washes (1× SSC, 0.5× SSC, 0.1× SSC twice each for 15 min) and blocking with culture medium (DMEM, 20% FCS, 0.5% BSA), sections were incubated with rabbit anti-dig-HRP Ab (1/100) in buffer (100 mM Tris-HCl, 150 mM NaCl (pH 7.5) with 1% rabbit serum) or with anti-dig alkaline phosphatase for riboprobes, overnight at 4°C. For oligoprobes, avidin-biotin block (Vector Laboratories) was followed by signal amplification using the Genpoint tyramide kit, and color was developed with diaminobenzidine (Vector Laboratories). For riboprobes, lamivudine block was performed, and 5-bromo-4-chloro-3′-indolyphosphate p-toluidine salt was used as the color substrate. Sections were briefly counterstained in Meyer’s hematoxylin (Gurr) or nuclear Fast Red (Sigma-Aldrich) for riboprobes and then mounted in aqueous medium and dried in an incubator before microscopic analysis.

The Leica laser microdissection microscope (ASLMD) was used to excise cell groups of histological tissue sections. Conventional microscopy slides were coated with a synthetic membrane held in place by a gene frame. Membrane-coated slides were then UV irradiated to destroy RNases. Snap frozen liver samples were mounted in OCT, and 10-μm sections were cut using a cryostat onto the membrane side of the slides. Tissue sections were fixed briefly in 70% ethanol, before nuclear counterstaining with Meyer’s hematoxylin (Gurr). For dehydration, the sections were immersed for 30 s each in 70% ethanol, 95% ethanol, and for 60 s in absolute ethanol. Processed slides were stored in a desiccator containing silica gel for at least 3 days before LCM. Tissue of interest was cut and captured directly into 50 μl of guanidine thiocyanate-containing lysis buffer plus 2-ME (0.7 μl/100 μl of lysis buffer). Visualization and cutting were conducted at ×200 magnification. Previous work determined that at least 1 × 10⁶ μm² of excised tissue was required for reliable RNA extraction and cDNA synthesis.

The Absolutely RNA Microprep kit (Stratagene) was used for RNA extraction according to manufacturer’s instructions. First-strand cDNA was synthesized from 7.7-μl RNA samples using random hexamers and Impromptu Reverse Transcriptase (Promega). cDNA was used for real-time RT-PCR as described above. We controlled for the area of tissue dissected (1 × 10⁶ μm²) and used RT-PCR for β-actin to assess integrity of extracted RNA before real-time mRNA quantitation with normalization to 18S rRNA.

Rat SAMs were detected in 4-μm paraffin sections using ED-1 Ab (1/150; Serotec) after citrate/microwave Ag retrieval as previously described (1). MMP13 was detected in rat liver by incubation with mouse monoclonal anti-MMP13 (1/80 dilution, clone VIII A2; Biocarta) after citrate/microwave Ag retrieval. After blocking endogenous peroxidase (hydrogen peroxide, 0.3% w/v in methanol), endogenous avidin-biotin and incubating with culture medium (DMEM, 20% FCS, 0.5% BSA), the Universal Quick Kit (Vector Laboratories) was used for Ag detection. Diaminobenzidine or Fast Red (Vector Laboratories) were used as chromogens, and counterstaining was with Meyer’s hematoxylin (Gurr).

H&E or picrosirius red were used for histochemical analysis of paraffin sections as previously described (19).

Rat SAMs in H&E-stained sections were counted in 10 randomly selected high power fields (HPF) by two independent, blinded observers. A positive cell was identified on the basis of characteristic morphology, a nucleus and brown intracellular pigment (ceroid) as evidence of phagocytosis (21) and included only if it was within or apposed to the scar. Counts were expressed as mean score ± SD.

For morphometric analysis of picrosirius red-stained collagen, 10 serial images from n = 3 control (olive oil) and n = 5–6 CCl₄-treated mmp13^−/− and WT mice were digitally captured in a blinded manner at ×200 magnification. The area of collagen staining was measured using Fovea Pro (Reindeer Graphics) (1), and results are expressed as mean ± SE and groups compared using an unpaired Student t test. p < 0.05 was considered statistically significant.

Rats, treated for 4, 6, 8, and 12 wk with CCl₄, developed liver fibrosis with increasing histopathological severity and diminishing degrees of reversibility. Indeed, remodeling of scar after 4 wk of injury led to a rapid restitution of normal or near-normal liver architecture, but 12 wk of injury induced an advanced micronodular cirrhosis from which incomplete resolution occurred (13, 19). To evaluate the potential role of MMP13 in the regression of liver fibrosis, we studied the expression profile of mmp13 mRNA using real-time RT-PCR during spontaneous resolution of the 12-wk CCl₄ rat cirrhosis model (Fig. 1 A). mmp13 transcript was up-regulated after cessation of CCl₄ injury, increasing 70-fold relative to untreated control liver at day 14 of resolution. Thereafter, mRNA expression declined significantly with time. This pattern of expression was consistent with our previous histological data, which showed that the majority of matrix remodeling had occurred by day 84 after cessation of injury (19).

There was a close relationship between the relative expression of mmp13 mRNA and the number of SAMs during resolution of injury in the rat model (Fig. 1, A and B), suggesting that SAMs might be a major source of MMP13.

We investigated the population dynamics of SAMs in our 12-wk model of advanced micronodular cirrhosis by direct counting of scar-related ceroid-laden cells with typical macrophage morphology, thus defining a highly phagocytic population of macrophages (Fig. 1, B and C). Although macrophages were numerous early in resolution, being most abundant at 14 days after cessation of injury, late recovery was characterized by a paucity of macrophages and incomplete remodeling of fibrotic scars even after 366 days of spontaneous resolution. These findings suggest that the presence of SAMs may be a requirement for complete histological resolution of advanced fibrosis. Indeed, in the 4-wk rat CCl₄ injury model where complete remodeling of fibrosis occurs, SAMs persist throughout the brief recovery period with an average of 13 cells per HPF remaining after 28 days of spontaneous regression (1, 13).

Quantitation of whole liver mRNA gives little information regarding topographical variations in gene expression. To study the hepatic scars and adjacent milieu, we used LCM to select areas of hepatic scarring and areas of unscarred parenchyma (Fig. 2, A and B). This technique allowed location specific quantitation of gene expression. Using real-time RT-PCR, we observed that mmp13 mRNA was restricted to areas of resolving fibrosis, rich in SAMs (Fig. 2 C).

The hepatic scar and its interface constitute a heterogeneous cell population. To establish the cellular source of mmp13, we used in situ hybridization in a murine CCl₄ model of liver fibrosis. We used a short oligonucleotide probe and, in addition, two ∼400-bp RNA riboprobes to reinforce the specificity of our findings. By both methods, we identified a population of large rounded scar-associated phagocytes as the principal source of mmp13 (Fig. 3, A–E), but there was no evidence that HSCs or myofibroblasts contained mmp13 transcripts. Moreover, SAMs were well placed to facilitate matrix degradation, lying within and apposed to fibrotic scars. HSCs did, however, express abundant timp1 (Fig. 3,H), high levels of which suppress local MMP activity during progressive fibrosis and inhibit apoptosis of activated myofibroblasts (13). Note the morphological distinction between the SAMs producing mmp13 transcript (Fig. 3, A, C, and D) and the HSCs producing timp1 (Fig. 3 H).

Having detected mmp13 mRNA in cells that resembled macrophages morphologically, we immunolocalized the cellular protein. MMP13 protein was detected in cells with histological features of SAMs (Fig. 3,F) and having similar topography to the population of phagocytes observed using in situ hybridization. Dual immunostaining using the cytoplasmic pan-macrophage marker ED-1 (the rat equivalent of human CD68) confirmed that MMP13-positive cells were SAMs (Fig. 3 G) in rat liver tissue. There are no anti-mouse MMP13 Abs that identify the protein in tissue sections. Eighty-three percent of ED-1-positive SAMs costained for MMP13 at day 14 of resolution in the 12-wk rat CCl₄ model. Importantly, immunohistochemistry revealed that apart from occasional spindle-shaped cells, there were no other discernable cell types within the liver with consistent MMP13 positivity.

We used a conditional macrophage ablation mouse model (CD11b-DTR) to study the effect of depletion of SAMs on hepatic mmp13 transcript level after CCl₄ injury (1). This model uses the low affinity of mouse DT receptor (heparin-binding epidermal growth factor receptor) compared with the human receptor. Transgenic expression of the human DT receptor confers sensitivity in mice only in macrophages and permits ablation in vivo after toxin injection. The CD11b-DTR construct uses the CD11b gene-regulatory elements for macrophage-specific expression. Using this model, monocytes in the circulation are depleted by 77 ± 14% (22) following DT, and macrophages in the liver scars are depleted by 78 ± 9% (1). In transgenic mice depleted of SAMs, we observed a 5-fold reduction in mmp13 transcript level (p = 0.005) by real-time RT-PCR of whole liver extracts (Fig. 4,A). The number of mmp13-positive cells was also decreased (p < 0.05) by in situ hybridization (Fig. 4 B).

We induced advanced liver fibrosis with 4-wk CCl₄ injury in cohorts of MMP13-deficient and WT mice and allowed them to recover spontaneously for up to 5 days. Livers were harvested at peak fibrosis (24 h after final injection of CCl₄) and at day 1, day 3, and day 5 after cessation of injury. Analysis of H&E-stained sections at peak fibrosis revealed equivalent necroinflammatory injury in mmp13^−/− and WT mice (Fig. 5, A and B). Immunohistochemistry for infiltrating neutrophils and macrophages showed no differences in hepatic inflammatory cell recruitment. At peak fibrosis, the mean number of neutrophils counted per HPF in the livers of mmp13^−/− mice (65 ± 8.6) was similar to that of WT mice (57.6 ± 3.3). Furthermore, CD68⁺ SAM numbers per 10 HPF in mmp13^−/− mice (3.56 ± 0.48) were not statistically different to WT mice (3.12 + 0.22) at peak fibrosis and throughout the recovery period (data not shown).

After picrosirius red staining for collagen, extensive and comparable scarring was observed at peak fibrosis in both mmp13^−/− and WT livers but not in vehicle (olive oil) controls (Fig. 5, C and D). However, remodeling of fibrosis was retarded in MMP13-deficient mice throughout the period of resolution. Quantitation of the picrosirius red area of staining by morphometry at day 5 of recovery (p < 0.05) confirmed the retarded remodeling phenotype (Fig. 5,G), where residual septal bands and perisinusoidal fibrosis persisted in mmp13^−/− livers in contrast to the WT livers where only minimal scarring remained (Fig. 5, E and F).

MMP13 deficiency did not abrogate regression of fibrosis completely, only retarded regression. One possible explanation is that other MMPs might take on additional roles in this mouse model. Real-time RT-PCR was used to determine whether mmp13 gene disruption had resulted in compensatory alterations in the expression of other candidate collagenases (Table I). Although there were trends toward increased mRNA expression of mmp2, mmp8, and mmp14 in mmp13^−/− mice relative to WT animals, the differences were not statistically significant.

For the first time, we have shown clear evidence that SAMs are a major source of MMP13 in resolving liver fibrosis. The observed phenotype in MMP13-deficient mice suggests that this collagenase plays a role in macrophage-directed remodeling of scar. Macrophages have been shown to promote resolution of liver fibrosis (1) and regulate repair processes in other organs (23, 24, 25, 26, 27). We hypothesized that SAMs were a potential source of hepatic MMP13 during regression of advanced fibrosis. To determine its expression profile, cellular source, and relative importance in vivo in mediating scar remodeling, we have used a series of distinct but complementary animal models.

Previous experiments using CCl₄ models of rat liver fibrosis demonstrate that SAMs populate hepatic scars during injury and repair (1). The number of SAMs increased in proportion to the duration of CCl₄ treatment, but they were also associated with resolving scar during spontaneous resolution. In the current study, hepatic MMP13 protein expression also increased as the period of CCl₄ injury was extended from 4 through 12 wk (our unpublished observation). This temporal correlation between SAM accumulation and MMP13 production was even more pronounced during regression of fibrosis. After 12 wk of CCl₄, we observed a prolonged increase in hepatic mmp13 transcript above peak fibrosis levels extending to 84 days of recovery, suggesting that MMP13 was actively engaged in the remodeling process. Marked up-regulation was seen as early as 3 days after cessation of injury, with maximal mRNA expression being at day 14 of resolution (Fig. 1 A). Furthermore, once SAMs had been lost from the recovering liver, further degradation of the fibrotic bands was minimal. These results differ from our earlier findings in the 4-wk fully reversible rat CCl₄ model, where mmp13 transcript level was increased at peak fibrosis and remained elevated during 28 days of resolution (13). SAMs were apparent throughout the remodeling phase, suggesting that their presence (and production of MMP13) might be a critical determinant of complete histological restitution.

Macrophages are recognized to be a major source of MMPs (2, 3) and can influence matrix turnover both directly by producing MMPs (24) and indirectly by secretion of cytokines including IL-1 and TNF-α that modulate MMP production by other resident cell types (25). We have provided compelling evidence, by both in situ hybridization (Fig. 3, A–E) and immunohistochemical staining (Fig. 3,F), that macrophages represent the primary source of MMP13 in relevant models of hepatic fibrosis; findings that were later confirmed using macrophage-specific markers (Fig. 3 G). These macrophages were seen almost exclusively within areas of scarring (SAMs) and therefore were ideally situated to facilitate matrix degradation. Moreover, when areas of fibrosis and unscarred liver tissue were separately dissected out by LCM, we detected mmp13 mRNA only in and around the scars.

The CD11b-DTR-transgenic mouse model provided a robust system to selectively deplete the cell population of interest (SAM) and observe any downstream effect on mmp13 transcript level and resolution of fibrosis. Our data show that macrophage depletion resulted in a 5-fold reduction in mmp13 transcript level by real-time RT-PCR, results that were consistent with subsequent experiments using in situ hybridization and counting of mmp13-positive cells, supporting the hypothesis that SAMs represent a major (but not exclusive) cellular source. DT administration did not abolish hepatic MMP13 entirely, a possible explanation being that macrophage depletion is not absolute in this model (1).

It is plausible that other cells capable of expressing MMPs may also contribute to matrix degradation by releasing collagenolytic enzymes into the pericellular milieu. Previous studies have documented transient expression of MMP13 by HSCs activated by primary culture on plastic (13) or after TNF-α (18) or halofuginone (28) administration, but the relevance of these observations to HSCs in vivo remains uncertain. More recently, Watanabe et al. (15) also suggested that HSCs are a source of MMP13. However, earlier work using 4- and 8-wk CCl₄ resolution models, together with the data in this study, demonstrate that MMP13 expression does not correlate with the number of activated HSCs and myofibroblasts, whereas TIMP-1 and collagen-1 decrease in proportion to the loss of activated HSCs from the liver (13, 19). Indeed, using in situ hybridization in serial sections of CCl₄-treated mouse liver, we detected abundant timp1 transcript but only rare mmp13 in scar-associated spindle cells. These data concur with previous studies which have identified HSCs as the primary source of inhibitory TIMPs in human liver disease (29) and in animal models of liver fibrosis both in vitro in cell culture models of HSC activation (30) and in vivo using in situ hybridization and immunohistochemistry (31). However, in our rodent models of liver fibrosis and spontaneous regression, HSCs are not a major source of MMP13.

The effect of MMP13 gene knockout has recently been studied in several animal models of tissue turnover and injury-repair in different organs. These studies have demonstrated that MMP13 plays an important role in bone development and remodeling and in collagen breakdown within atherosclerotic plaques (20, 32, 33), but not in epidermal wound healing (34). In a biliary fibrosis model of progressive liver injury, MMP13 deficiency attenuated hepatic inflammation and fibrosis (35). However, bile duct ligation is a complex model involving proliferation of stem cell components in association with myofibroblasts, whereas CCl₄ is a model of inflammation, repair and resolution. We have shown that MMP13 deficiency reduced the capacity of the liver to remodel scar by 5 days of recovery in a mouse model of advanced liver fibrosis, although gene deletion did not inhibit resolution completely. Chronic CCl₄ toxicity induced inflammatory and fibrotic liver injury that was similar in gene-deleted and WT mice such that our observations during the resolution phase were directly comparable. Interestingly, a perisinusoidal associated fibrosis persisted in mmp13^−/− livers. This distribution of fibrillar collagen has been shown to be especially rich in SAMs after CCl₄ injury (1). The likely explanation for our mmp13^−/− phenotype is partial compensation by other MMPs capable of degrading interstitial collagens (MMP8, MMP14, MMP2). Although for each individual MMP the trend toward up-regulation was not statistically significant in mmp13^−/− mice relative to WTs (Table I), in combination these proteases might materially contribute to matrix degradation. Both MMP2 and MMP14 have been detected in animal models of liver fibrosis and in chronic hepatitis and cirrhosis in humans, with mRNA localizing to HSCs (36, 37, 38). They are also up-regulated during resolution of fibrosis, although the actions of MMP2 appear to be complex and somewhat paradoxical in that this enzyme may contribute to fibrogenesis via proproliferative effects on HSCs (32).

In contrast to humans, in whom MMP1 is expressed as the major interstitial collagenase, the ortholog of MMP1 (McolA) is expressed in mice at low levels (39). Instead, MMP13 is widely believed to remodel fibrillar collagen in rodent tissues. Despite structural and functional similarities, MMP1 and MMP13 show contrasting responses to cytokines and growth factors and have distinct MMP-dependent activation mechanisms (7). Therefore, one should not automatically assume that observations from rodent models of liver fibrosis are germane to human disease. Nevertheless, in preliminary studies using immunohistochemistry, we have demonstrated MMP13 protein expression in two separate cases of human alcoholic cirrhosis within large scar-associated phagocytes, consistent with our rodent studies (data not shown). As well as cleaving collagen-1 at a single locus in the triple helix, MMP13 cleaves collagen-1 at a site in the N telopeptide which destabilizes cross-linked collagen. This may facilitate the breakdown of mature, highly cross-linked collagen in tissues that have been rendered resistant to MMP1.

The identity and source of the rodent collagenase that mediates matrix remodeling in liver fibrosis has proved elusive, despite the importance of these data to the interpretation of models designed to determine the pathogenesis of fibrosis and provide proof of concept of potential therapies. The identification of SAMs as the primary source of interstitial collagenase (at least in rodents) represents a major advance in our understanding of fibrosis resolution. Our preliminary data in human cirrhosis indicate that MMP13 might also play a more prominent role in human liver fibrosis than was previously considered.

We thank Dr. Stephen Krane (Harvard Medical School, Boston, MA) for the gift of MMP13 knockout mice and Dr. Joseph Bonventre (Harvard Medical School) and Dr. Stuart Forbes (University of Edinburgh, Edinburgh, U.K.) for assistance.

J. P. Iredale is a consultant for GEC Healthcare advising on noninvasive diagnosis of fibrosis.