Background and Aims:
Hepatic steatosis has been shown to be associated with lipid peroxidation and hepatic fibrosis in a variety of liver diseases including non-alcoholic fatty liver disease. However, the lobular distribution of lipid peroxidation associated with hepatic steatosis, and the influence of hepatic iron stores on this are unknown. The aim of this study was to assess the distribution of lipid peroxidation in association with these factors, and the relationship of this to the fibrogenic cascade.
Liver biopsies from 39 patients with varying degrees of hepatic steatosis were assessed for evidence of lipid peroxidation (malondialdehyde adducts), hepatic iron, inflammation, fibrosis, hepatic stellate cell activation (α-smooth muscle actin and TGF-β expression) and collagen type I synthesis (procollagen α1 (I) mRNA).
Lipid peroxidation occurred in and adjacent to fat-laden hepatocytes and was maximal in acinar zone 3. Fibrosis was associated with steatosis (P < 0.04), lipid peroxidation (P < 0.05) and hepatic iron stores (P < 0.02). Multivariate logistic regression analysis confirmed the association between steatosis and lipid peroxidation within zone 3 hepatocytes (P < 0.05), while for hepatic iron, lipid peroxidation was seen within sinusoidal cells (P < 0.05), particularly in zone 1 (P < 0.02). Steatosis was also associated with acinar inflammation (P < 0.005). α-Smooth muscle actin expression was present in association with both lipid peroxidation and fibrosis. Although the effects of steatosis and iron on lipid peroxidation and fibrosis were additive, there was no evidence of a specific synergistic interaction between them.
These observations support a model where steatosis exerts an effect on fibrosis through lipid peroxidation, particularly in zone 3 hepatocytes.
Hepatic steatosis is associated with increased hepatic fibrosis in a number of liver diseases, particularly non-alcoholic fatty liver disease, chronic hepatitis C infection and alcoholic liver disease.1–8 This increased fibrosis appears to be a consequence of hepatic stellate cell activation and the synthesis of collagen type I, the major collagen in fibrotic but not normal liver tissue.9
Lipid peroxidation and exposure to products of lipid peroxidation are an important mechanism of hepatic stellate cell activation in vitro, resulting in increased procollagen α1 (I) gene expression.10–12 The association between lipid peroxidation and hepatic fibrosis has been shown for a variety of liver diseases including hemochromatosis, alcohol-induced liver injury and chronic hepatitis C.3,4,6,13–15
The evidence linking hepatic steatosis with lipid peroxidation and stellate cell activation is not as clearly defined. Animal studies have demonstrated that lipid peroxidation occurs in both acute and chronic steatosis, and is induced by a variety of agents.16,17 In alcoholic liver disease, hepatic stellate cell activation correlates with the severity of steatosis, with products of lipid peroxidation co-localizing with steatosis and fibrosis.6,15 In hepatitis C infection, steatosis is associated with lipid peroxidation,4 with lipid peroxidation being present in areas of active fibrogenesis.3 The relationship between lipid peroxidation and hepatic stellate cell activation in relation to hepatic steatosis has not been assessed, although it is presumed, as in other liver diseases, that these factors are linked.18,19 Moreover, the lobular and cellular distribution of lipid peroxidation in relation to steatosis has not been described and may be relevant to the pattern of fibrosis seen in these patients.
There is increasing interest in the role of iron as a cofactor for fibrosis in a number of liver diseases. Increased iron has been shown in both human studies and animal models to be associated with lipid peroxidation and hepatic stellate cell activation.3,4,13,20–22 We have recently shown that: (i) hepatic iron stores are increased in some patients with non-alcoholic steatohepatitis, principally caused by the presence of mutations of the hemochromatosis gene, HFE; and (ii) that both hepatic steatosis and iron are risk factors for fibrosis in these patients.1 Although this fibrosis is presumably related to lipid peroxidation and stellate cell activation, this has not been shown, and the interaction between fat and iron in the development of hepatic fibrosis is unknown.
The aims of this study were: to determine the relationship between hepatic steatosis, lipid peroxidation and fibrosis; to assess the lobular and cellular distribution of lipid peroxidation in relation to steatosis; and to determine if there is evidence of a synergistic interaction between hepatic steatosis and iron stores on lipid peroxidation and hepatic fibrosis. To examine these factors, liver biopsy tissue from 39 patients was assessed with respect to the grade of steatosis, hepatic iron stores, and evidence of lipid peroxidation and fibrosis. The lobular distribution of lipid peroxidation was assessed for both hepatocytes and sinusoidal cells within each acinar zone by using a semiquantitative scoring system. Multivariate logistic regression was used to assess the interactions between hepatic steatosis, lipid peroxidation and other histological variables.
The study population was drawn from our databases of individuals with non-alcoholic hepatic steatosis and HFE-related iron overload. Patients attended liver clinics at the Royal Brisbane Hospital and were included in the study if: (i) there was sufficient clinical and serologic data to exclude other significant liver disease including alcoholic liver disease; and (ii) there was sufficient tissue for the immunohistochemical and in situ hybridization studies detailed below. Thirty-three patients met these requirements (27 males, six females; aged 18–68 years, median age 43 years). Twenty-four patients had hepatic steatosis, with 13 of these having increased hepatic iron stores. Nine biopsies were from patients with mildly increased hepatic iron stores (median hepatic iron concentration 80 μmol/g dry weight, range 56–113 μmol/g dry weight) and no steatosis, allowing us to assess the effect of mild iron overload independent of steatosis and include this in our statistical modeling. Six biopsies from donor liver tissue with neither steatosis nor stainable iron stores (four males, two females; aged 14–70 years, median age 36 years) formed our control group.
No patient consumed more than 60 g of alcohol per day, 36 drinking < 20 g/day, and three patients drinking between 20 and 60 g/day. No patient had chronic hepatitis B, while chronic hepatitis C was excluded in all but seven cases where the biopsies were collected prior to the availability of serologic testing and the patients lost to follow up. This study had the approval of the Royal Brisbane Hospital and Princess Alexandra Hospital Research Ethics Committees.
Hepatic iron stores and HFE gene testing
Hepatic iron concentration (HIC) was available in 31 patients and ranged between 1 and 352 μmol/g dry weight of liver. The HIC was < 40 μmol/g dry weight in 12 patients, between 40 and 80 μmol/g dry weight in nine patients, between 80 and 200 μmol/g dry weight in seven patients, and > 200 μmol/g dry weight in the remaining three patients. The HFE gene test was available on 27 patients. Eight were homozygous normal, one was heterozygous for the His63Asp mutation, three were heterozygous for the Cys282Tyr mutation, two patients were Cys282Tyr/His63Asp compound heterozygotes and 13 patients were homozygous for the Cys282Tyr mutation of HFE. Of the remaining 12 patients where HFE gene testing was not available, a further four had a diagnosis of hemochromatosis made on clinical grounds as described previously,1 while eight patients had no stainable iron and a maximum HIC of 26 μmol/g dry weight. Three patients with hemochromatosis and steatosis had postphlebotomy biopsies available for comparison. Data from these three biopsies were not used in the statistical analysis.
Biopsy sections were stained with hematoxylin and eosin for morphologic evaluation, Perls’ Prussian Blue stain for assessment of iron loading, and Gordon and Sweet’s reticulin, Masson’s trichrome or hematoxylin Van Gieson stains for assessment of fibrosis.
Immunohistochemistry was carried out for adducts of malondialdehyde (MDA), a product of lipid peroxidation; α-smooth muscle actin (α-SMA), a marker of hepatic stellate cell activation; and transforming growth factor (TGF)-β, a profibrogenic cytokine involved in fibrogenesis. Collagen type I is the principal collagen synthesized by activated hepatic stellate cells, and in situ hybridization was performed for procollagen α1 (I) mRNA. The staining in these sections was scored as outlined below, and assessed in relation to hepatic iron stores, steatosis, acinar and portal inflammation and fibrosis.
Immunohistochemistry for malondialdehyde adducts, α-smooth muscle actin and transforming growth factor-β
Monospecific antiserum to MDA-lysine adducts were generated by using previously described techniques.21,23 The resultant antiserum is specific for MDA-lysine adducts and reacts with these adducts on a variety of different proteins. Sections were immunostained as previously described with the specific MDA antibody at a dilution of 1:250.21,22,24
α-Smooth muscle actin (α-SMA) was detected with a mouse monoclonal anti-α-SMA antibody (clone 1A4; Sigma Chemical Company, St Louis, MO, USA) as previously described.20 For TGF-β1 immunohistochemistry, sections were stained with an anti-TGF-β1,2,3 monoclonal antibody (Genzyme Corporation, Cambridge, MA, USA), as previously described.25
In situ hybridization for procollagen α1 (I) mRNA
Human procollagen α1 (I) digoxigenin labeled riboprobes (1.5 kb) were produced by using a standard RNA labeling reaction, and then a 300-bp fragment was produced by alkaline hydrolysis.25 Hybridization and the detection of bound riboprobe was performed as previously described.25,26 Sections underwent immunohistochemistry for detection of α-SMA to co-localize procollagen α1 (I) mRNA to activated hepatic stellate cells.
The degree of inflammation and fibrosis was assessed and graded according to the method of Scheuer.27 As in our previous study, steatosis was graded 1 (< 30% hepatocytes affected), 2 (30–70% hepatocytes affected) and 3 (> 70% hepatocytes affected), while the Perls’ stain for iron was graded 0–4.1
The emphasis of the present study was on the distribution of lipid peroxidation in relation to hepatic steatosis, and so this could be statistically assessed; the MDA staining of all cases was reviewed and the following semiquantitative scoring system devised. The extent of MDA staining of connective tissue in portal tracts was graded as follows: (+) scattered irregular staining sometimes confined to the cytoplasm of connective tissue cells; (++) diffuse staining of 33–66% of connective tissue; (+++) diffuse staining of 67–100% of connective tissue. For sinusoidal cells in Rappaport zones 1–3 the MDA-adduct scoring was as follows: periportal and acinar zone 1, (+) fine short extensions of staining between periportal hepatocytes or single staining dendritic shapes in sinusoids, (++) short broad extensions of staining between periportal hepatocytes often with minor loss of periportal hepatocytes, (+++) multiple broad longer extensions of staining between periportal hepatocytes or more extensive networks of staining in subsinusoidal areas; acinar zones 2 and 3, (+) staining associated with small inflammatory foci or confined to small dendritic areas in subsinusoidal space, (++) staining associated with scattered larger inflammatory foci, (+++) staining associated with grouped larger inflammatory foci and/or extensive networks of pericellular and subsinusoidal staining. The extent of MDA staining of hepatocytes in Rappaport zones 1–3 was graded as follows: (+) small numbers of isolated hepatocytes; and (++) line of hepatocytes in the immediate periportal or perivenular layer with or without scattered cells elsewhere.
α-Smooth muscle actin expression was graded 1–4 by using a previously described semiquantitative scoring system.28 Tumor growth factor-β immunohistochemistry was graded by using a similar semiquantitative scoring system to α-SMA: (+) for small numbers of either isolated sinusoidal cells or hepatocytes; (++) for clusters of either sinusoidal cells or hepatocytes in either the periportal or perivenular regions of the acinus; (+++) for periportal or perivenular sinusoidal cells or hepatocytes bridging into zone 2; and (++++) for pan lobular distribution in sinusoidal cells or hepatocytes.
The MDA-adduct immunohistochemistry scored as above was assessed with respect to the grades of steatosis, fibrosis, inflammation and Perls’ stain as described. A similar analysis was carried out for α-SMA and TGF-β immunohistochemistry and procollagen α1 (I) mRNA in situ hybridization. Associations to discrete ordinal variables were tested for significance by using the OLOGIT procedure of the STATA statistical package (STATA Corporation, College Station, TX, USA), which performs multivariate logistic regression allowing for linear modeling of ordinal variables having more than two levels. This method was chosen because of the discrete categoric nature of the response variables. Test statistics were likelihood ratio and chi-squared tests, which measure the change of deviance accompanying removal of explanatory variables from the linear model. Where a statistical test involved the removal of a single explanatory variable from the linear model, standard one-sided t-tests on the regression coefficient for that variable were used instead as the test statistic. Throughout, P values less than 0.05 were considered significant.
Products of lipid peroxidation were detected around central veins and were obvious in both sinusoidal cells and hepatocytes (Fig. 1a). Lipid peroxidation in association with hepatic steatosis was also seen scattered throughout the hepatic acinus often, but not exclusively, in association with inflammatory foci (Fig. 1b). However, not all lipogranulomas and inflammatory foci had evidence of lipid peroxidation. Additionally, MDA adducts co-localized with hepatic iron stores, with MDA adducts present in zone 1 predominantly in sinusoidal cells at the edge of the acinus (Fig. 1c).
α-Smooth muscle actin staining in relation to steatosis was present adjacent to central veins and was most intense in association with lipogranulomas (Fig. 2). In patients with iron overload, α-SMA expression followed the expected pattern, being intense in zone 1. As reported previously,20 phlebotomy was associated with a generalized reduction in peri-sinusoidal hepatic stellate cell expression of α-SMA, except in the areas adjacent to lipogranulomas (Fig. 2c,d).
Analysis of the relationship between lipid peroxidation and fibrosis
Steatosis and iron stores (Perls’ grade) were both significantly associated with fibrosis (P < 0.005 and P < 0.002, respectively; Table 1). This confirmed our previous observations,1 although 14 biopsies in the present study were used in the earlier report. Additionally, MDA staining was significantly associated with fibrosis (P < 0.05). When MDA staining was included in the model, the strength of the association of steatosis and iron stores with fibrosis weakened (P < 0.04 and P < 0.02; respectively, Table 1). This is consistent with a model, where at least part of the effect of steatosis and hepatic iron stores on fibrosis is mediated through lipid peroxidation. Neither age nor gender was associated with any of variables assessed.
|Perls’ grade||< 0.005||< 0.04|
|Steatosis||< 0.002||< 0.02|
|Portal inflammation||< 0.002||0.005|
|Portal tracts||–||< 0.07|
|Sinusoidal cells-zone 1||–||< 0.004|
|Sinusoidal cells-zone 2||–||NS|
|Sinusoidal cells-zone 3||–||NS|
|Hepatocytes-zone 3||–||< 0.10|
When MDA staining was examined by cell type, fibrosis was associated with evidence of lipid peroxidation in sinusoidal cells (P < 0.05), with the strongest association between sinusoidal cell MDA and fibrosis being seen in zone 1 (P < 0.004). There was a trend towards an association between hepatocyte MDA staining and fibrosis (P < 0.10), particularly in zone 3 (P < 0.10); however, this was not statistically signifi-cant. There was no evidence of a synergistic interaction between steatosis and hepatic iron stores on lipid peroxidation or fibrosis.
Zonal distribution of lipid peroxidation in relation to steatosis and iron stores
When examined on a zone by zone basis, steatosis was associated with evidence of lipid peroxidation in hepatocytes in zone 3 (P < 0.05, Table 2). Raised hepatic iron stores (Perls’ grade) were associated with evidence of lipid peroxidation in sinusoidal cells (P < 0.05), with this effect being strongest for sinusoidal cells in zone 1 (P < 0.02, Table 2). There was no significant association between iron stores and the presence of MDA adducts in hepatocytes when these were considered as a group; however, when examined on a zone by zone basis, there was an association between iron and lipid peroxidation for hepatocytes in zone 2 (P < 0.006, Table 2).
|Hepatic iron stores||Steatosis|
|Acinar inflammation||NS||< 0.005|
|Sinusoidal cells-zone 1||< 0.02||NS|
|Sinusoidal cells-zone 2||NS||NS|
|Sinusoidal cells-zone 3||NS||NS|
|Hepatocytes-zone 2||< 0.006||NS|
|Hepatocytes-zone 3||NS||< 0.05|
Portal and acinar inflammation
Portal inflammation, but not acinar inflammation was associated with fibrosis (P < 0.002, Table 1). Acinar inflammation was related to steatosis (P < 0.005, Table 2), while portal inflammation appeared to be related to hepatic iron stores (P = 0.05, Table 2). Although on examining the biopsies lipid peroxidation appeared to occur in association with acinar inflammation, there was no statistically significant association with zonal distribution or cell-type MDA staining and acinar inflammation.
α-Smooth muscle actin and transforming growth factor-β1
When all 39 biopsies were studied, there was no correlation between α-SMA staining and either histological fibrosis or MDA staining. This appeared to be a consequence of an unexpected increase in expression of α-SMA in the donor liver tissue used as controls, despite these tissues having minimal steatosis, stainable iron and inflammation. Three of the six biopsies from donor livers had grade 3 α-SMA staining, and two biopsies had grade 2 α-SMA staining. When the six biopsies from donor livers were excluded from the analysis, significant associations were seen between lipid peroxidation and α-SMA (P < 0.001), and α-SMA and fibrosis (P < 0.04) in our subjects with hepatic steatosis and/or increased iron stores. In this analysis the association between lipid peroxidation and α-SMA expression was significant for both sinusoidal cell and hepatocyte MDA (P < 0.005 and P < 0.025, respectively).
Transforming growth factor-β staining was seen in both hepatocytes and sinusoidal cells. Overall, the intensity of TGF-β staining was low, with only four cases having grade 2 and one case grade 3 staining, despite good staining in our positive control tissue. The problems with significant α-SMA staining in donor liver tissue controls were also seen for TGF-β immunohistochemistry, with two of the donor liver biopsies having grade 2 TGF-β staining.
Procollagen α1 (I) mRNA
Only 30 biopsies had sufficient tissue for in situ hybridization for procollagen α1 (I) mRNA. There was minimal procollagen α1 (I) mRNA detected overall, 17 of the 30 biopsies studied had none and 11 of the remaining 13 cases had procollagen α1 (I) mRNA expressed in expanded portal tracts and occasional scattered sinusoidal cells.
The most striking observation from the present study in terms of understanding the events that lead to the perivenular and subsinusoidal fibrosis seen in hepatic steatosis, is the association between steatosis and lipid peroxidation in zone 3 hepatocytes. In a micropig model of alcoholic liver disease, MDA adducts were seen in zone 3, co-localized with areas of steatosis.15 More recently, Niemela et al. have shown that human subjects with alcoholic liver disease have evidence of lipid peroxidation in zone 3, although the cell types involved and the relationship of this to steatosis was not commented on.29 Additionally, Niemela et al. detected lipid peroxidation in patients with non-alcoholic fatty liver disease, although this was less than that seen in patients with alcoholic liver disease; the authors did not comment on its zonal distribution.29 The reason for the localization of MDA adducts to zone 3 may relate to the lower pO2 in this zone and associated oxidative stress. In patients with alcoholic liver diseases, hepatic stellate cell activation correlates with steatosis and is greatest in perivenular regions.6 The lipid peroxidation that occurs in zone 3 in both alcoholic and non-alcoholic steatosis is likely to be causally related to perivenular fibrosis.
We have previously shown that acinar inflammation is a risk factor for fibrosis in non-alcoholic steatohepatitis.1 In the present study, although lipid peroxidation appeared to be associated with acinar inflammation, this was not statistically significant. One explanation for this is that the inflammatory foci had a patchy distribution throughout the hepatic lobule. This heterogeneity would influence the studies’ ability to detect a significant association between steatosis and lipid peroxidation. There is evidence that inflammation can be associated with lipid peroxidation as products of lipid peroxidation have been shown to be neutrophil chemotactants,30 while in co-culture experiments, reactive oxygen species released by neutrophils result in lipid peroxidation and stimulate collagen synthesis by hepatic stellate cells.31
Increased hepatic iron stores were associated with evidence of lipid peroxidation in sinusoidal cells, most significantly for those sinusoidal cells in zone 1. This corresponds to the zonal distribution of iron. The sinusoidal cells that stained for MDA adducts were, based on their morphology and distribution, hepatic stellate cells (Fig. 1c). The distribution of MDA adducts in sinusoidal cells appears paradoxical as the major stimulus for lipid peroxidation in this situation (hepatic iron) is in the hepatocytes. In a previous study, Bedossa et al. exposed co-cultures of hepatocytes and hepatic stellate cells to carbon tetrachloride, and demonstrated strong staining for products of lipid peroxidation in hepatocytes and also in those hepatic stellate cells immediately adjacent to hepatocytes.12 When hepatic stellate cells were cultured on their own with carbon tetrachloride there was minimal activation. The present study and that by Bedossa et al.,12 provide evidence that products related to lipid peroxidation are released by hepatocytes and result in the formation of MDA adducts within hepatic stellate cells.
The observation that lipid peroxidation in our patients with steatosis and/or increased hepatic iron stores is associated with α-SMA expression is consistent with previous studies reporting the link between lipid peroxidation and hepatic stellate cell activation.3,10–13,32 The high expression of α-SMA and TGF-β in the donor liver tissue is intriguing. This donor liver tissue was collected after liver harvesting and prior to reperfusion of the graft. These observations suggest that hepatic stellate cell activation in relation to the events associated with brain death and organ procurement can occur independent of lipid peroxidation. This observation has not been reported previously, but given the current interest in oxidative stress and its consequences for donor livers in the peri-transplant period would seem worthy of further study.
The low levels of TGF-β protein and procollagen α1 (I) mRNA expression are also intriguing, and raise the possibility that decreased matrix degradation rather than increased synthesis of collagen I may be responsible for fibrogenesis in relation to hepatic steatosis and iron.
In conclusion, this present study has shown for the first time that there is a statistically significant association between non-alcoholic steatohepatitis and lipid peroxidation, particularly in zone 3 hepatocytes. However, the association between lipid peroxidation and hepatic steatosis is less dramatic than for iron. While there was no statistically significant synergistic interaction in the present study between iron and steatosis with respect to lipid peroxidation, both factors can contribute to hepatic fibrosis. As expected, lipid peroxidation appears causally related to fibrosis. There are potentially important clinical implications from our findings: hepatic steatosis may be an independent factor contributing to fibrosis in a variety of liver conditions, while phlebotomy therapy may be beneficial in patients with hepatic steatosis and iron overload. These issues warrant further study.
Human procollagen α1 (I) cDNA was kindly provided by Dr Glenda Gobe, Department of Pathology, The University of Queensland. Drs Ramm, Crawford and Powell are supported by an Institute grant from the National Health and Medical Research Council of Australia to the Queensland Institute of Medical Research. Dr Houglum was supported by the Department of Veteran Affairs and the United Liver Foundation.