Megamitochondria with crystalline inclusions (MMC) have been previously described in nonalcoholic fatty liver; however, their distribution within hepatic zones is unknown. We sought to determine this distribution from the core liver biopsy specimens of 31 patients: 8 males and 23 females, age range 21 to 72 years. Twenty-nine showed evidence of nonalcoholic steatohepatitis (NASH) on biopsy with steatosis, inflammation, varying degree of fibrosis, ballooned hepatocytes, and Mallory hyaline, and two patients had cryptogenic cirrhosis thought to represent “burned out” NASH. Identified by transmission electron microscopy, the abundance of MMC was compared between low-stage (fibrosis stages 1 and 2) and high-stage (fibrosis stages 3 and 4) groups and between zones with or without difference in fibrosis stage. Regardless of stage, the MMC were distributed equally in all zones and were abundant similarly in low- and high-stage groups. This abundance did not correlate with the degree of oxidative stress (4-hydroxynonenal staining) or with the abundance of ballooned hepatocytes. Consistent with age as a risk factor for more severe disease, the median age for the low-stage group was significantly lower than that of the high-stage group (P = .003). In conclusion, in NASH, the MMC seem to be distributed randomly among zones and without variation in abundance, regardless of the fibrosis stage. The exact function of these structures remains to be defined. In this study, their presence did not seem to correlate with the light microscopic injury pattern represented by ballooned hepatocytes or degree of oxidative stress defined by immunostaining for 4-hydroxynonenal. (HEPATOLOGY 2004;39:1423–1429.)
More than 50 years have elapsed since the initial description of mitochondrial abnormalities and impaired oxidative phosphorylation in experimental steatosis.1 Morphologically, the most striking change is the development of mitochondrial swelling with formation of crystalline inclusions within the mitochondrial matrix. Although once described as “para”-crystalline, optical diffraction studies by Sternlieb and Berger2 demonstrated that the crystalline inclusions are true crystals, although their composition remains uncertain. Similar abnormalities have been described in Wilson disease,3 alcoholic steatohepatitis (where they were believed to be markers of early or less severe alcohol-related injury),4 and more recently in nonalcoholic fatty liver, where their presence has been correlated indirectly with oxidative stress and their absence noted when cirrhosis develops.5, 6
We previously estimated that megamitochondria with crystalline inclusions (MMC) are seen in 5% to 15% of hepatocytes in nonalcoholic steatohepatitis (NASH) patients and in 5% to 10% of the mitochondria within an afflicted cell.5, 7 Although not invariable, crystalline inclusions typically appear in close association with swollen mitochondria, which may be rounded or elongated. The crystals occur as long parallel strands, which often deform the shape of the mitochondrion. Each single crystalline strand is approximately 10 nm in diameter, and typically there are 20 nm between strands. The crystal strands usually appear as bundles viewed longitudinally or as an evenly spaced matrix in cross section (Fig. 1A, B). They often appear to displace the cristae and at times appear to have continuity with the cristae, suggesting a common origin.8
Although commonly seen in NASH patients, there is variation in the ease with which these structures can be observed by electron microscopy in a given biopsy. This could represent regional differences in local oxidative stress and disease activity within the hepatic zone. Alternatively, it is possible that this variation represents random distribution of these structures, as may be seen with a more diffuse injury. To determine whether these structures correlate with the injury pattern seen by light microscopy, we used electron microscopy approaches, including the mesa technique for trimming of embedded material, to assess the zonal distribution of the MMC from the liver biopsy tissues obtained in NASH patients with low- and high-stage fibrosis.
Liver biopsy specimens from 31 patients were studied. All except for two patients had evidence for NASH on biopsy, with steatosis, inflammation, varying degrees of fibrosis, ballooned hepatocytes, and Mallory hyaline. The remaining two patients with risk factors for NASH (obesity, diabetes mellitus type 2, and hypertriglyceridemia) had cryptogenic cirrhosis thought to represent “burned-out” NASH.9 There were 8 males and 23 females (age range, 21–72 years). None of the patients had serological or clinical findings of chronic viral hepatitis (B or C), primary biliary cirrhosis, primary sclerosing cholangitis, autoimmune hepatitis, alcohol- or drug-induced liver disease, Wilson disease, hemochromatosis, or α-1-antitrypsin deficiency. All of the subjects had risk factors for NASH, such as obesity (body mass index ≥ 30), diabetes mellitus type 2, hypertriglyceridemia, or a combination thereof (see Tables 1 and 2).
Table 1. Low-Stage NASH: Fibrosis Stages 1 and 2
MMC Zone 1
MMC Zone 2
MMC Zone 3
Abbreviations: BMI, body mass index; DM2, diabetes mellitus type 2; TG, hypertriglyceridemia; Balloon score, ballooned hepatocyte score; NA, specimen not available.
NOTE. For each subject, the numerical value recorded in each zone represents the mean score of the MMC of the semithin sections.
The zone was not evident on any of the sections examined.
Standard percutaneous liver biopsy was performed in all 31 patients using a 1.6-gauge Jamshidi needle, yielding 31 specimens with a core length of 2.5 to 3.5 cm. These liver biopsy specimens were judged to be adequate for analysis by our pathologist, based on the sample size and the number of portal tracts observed. For comparison, these specimens were categorized by an experienced liver pathologist (JCI) into two NASH groups of low stage (fibrosis stages 1 and 2) and high stage (fibrosis stages 3 and 4), using Brunt's classification.10 In this classification, stages are defined as follows: stage 1 as zone 3 with perisinusoidal fibrosis, stage 2 as zone 3 with sinusoidal fibrosis and zone 1 with periportal fibrosis, stage 3 as bridging fibrosis, and stage 4 as cirrhosis. These zones are the light-microscopic regions within an acinar that represent the microcirculation distribution in a model conceived by Rappaport.11 In this study, zone 1 and zone 3 are represented by the periportal and perivenular regions, respectively, whereas zone 2 is defined by the region of non-zone 1 and non-zone 3. There were 17 patients in the low-stage group and 14 patients in the high-stage group. Because agents that may have activity in NASH could interact with mitochondrial form and function, we further looked at the possible effects of these medications: 8 patients (47%) with low-stage fibrosis and 11 patients (79%) with high-stage fibrosis were taking antidiabetic agents, antihyperlipidemic agents, vitamin E, or ursodeoxycholic acid (see Tables 1 and 2).
From each biopsy core, two nonadjacent segments (2–4 mm) were separated quickly after tissue procurement and fixed (from 4 hours to 2 weeks at 24°C) in a solution containing 4.0 % (wt/vol) paraformaldehyde and 2.5 % (wt/vol) glutaraldehyde in 0.1 M phosphate buffer (pH 7.2) for transmission electron microscopy. The remaining tissue was submitted for routine processing and light microscopy.
After initial fixation, the two original segments for transmission electron microscopy morphometry were divided further, yielding four tissue segments representing different regions of each liver. These tissues were then fixed further for 1 hour in 1.0 % osmium tetroxide at 24°C, dehydrated in acetone, and embedded in epoxy resin. Semithin sections (0.5 μm in thickness) from four sequential levels (separated by at least 50 μm) then were prepared from each resin-embedded block to provide 16 nonoverlapping regions for analysis from each patient's specimen.
Each region represented in the semithin sections was visualized by light microscopy for characteristic landmarks and was categorized according to its zone. After each semithin sectioning, shallow trimming of the representative region (mesa technique; Fig. 2) was carried out, preserving deeper-lying portions of the tissue for subsequent sectioning while creating a block face small enough to facilitate ultrathin sectioning. Using a diamond knife, ultrathin sections (70–80 nm in thickness) from each mesa surface then were prepared, were collected on 200 mesh copper grids, were stained with lead citrate and uranyl acetate, and were observed for MMC in a JEOL 100-CX electron microscope (Japan Electron Optics Limited, Tokyo, Japan).
To quantify the abundance of the MMC within zones, these ultrathin sections were examined by two experienced electron microscopists (JAR, CAD) who were blinded to the clinical history and stage. Based on our prior experience,5, 7 a semiquantitative scale was used to score the abundance of the MMC present in each semithin section: a score of “+++” signified the identification of MMC within 2 minutes of observation; a score of “++” signified the identification of MMC between 2 and 5 minutes of observation; a score of “+” signified the identification of MMC between 5 and 10 minutes of observation; and a score of “0” signified no MMC seen after 10 minutes of searching. Most of the abnormal mitochondria with crystalline inclusions were enlarged, with either an elongated or rounded shape. When data from the ultrathin sections from all 16 regions for each patient's specimen had been obtained, the mean scores of the MMC for each zone were calculated and tabulated, as shown in Tables 1 and 2.
In addition, determination of the abundance of the ballooned hepatocytes by light microscopy was carried out in 31 biopsy specimens by an experienced pathologist (JCI), using a semiquantitative scale adapted from Brunt et al.10 This scale is from 0 to 3, with 3 as being most abundant (see Tables 1 and 2). Using immunostaining, the degree of oxidative stress was assessed in 30 specimens (one was unavailable) by the immunoreactivity for 4-hydroxynonenal (4-HNE) that was quantified on a semiquantitative scale of 1 to 4, with 4 being the strongest reactivity (see Tables 1 and 2).12
Because the scores of the MMC, the immunostaining for 4-HNE, and the abundance of the ballooned hepatocytes were not normally distributed, nonparametric statistical methods were applied. Comparisons of the scores among the three zones within a similar stage group and between two identical zones from a different stage group were achieved by applying the Friedman and Wilcoxon rank-sum tests, respectively. Without separation by fibrosis stage, comparisons of the abundance of the MMC between zone 1 and zone 3 and between combined zone (1 and 2) and zone 3 were obtained by the Wilcoxon rank-sum test. To compare the overall scores of the MMC further between low-stage and high-stage groups, after obtaining a mean score from all the zones, the Wilcoxon rank-sum test also was used. Similarly, this test was used to compare the abundance of the MMC between two groups of low (0–1) and high (2–3) ballooned hepatocyte scores and between two groups of low (1–2) and high (3–4) immunoreactivity for 4-HNE. The ballooned hepatocyte scores and the degree of immunostaining for 4-HNE between low- and high-stage groups also were compared using a similar statistical method. For other comparisons, such as age, BMI, and gender distribution, the Wilcoxon rank-sum and the Fisher's exact tests were used. Type I error was controlled at an α test of 0.05. All tests were reviewed by an experienced statistician (KOA).
The study protocol was reviewed and approved by the Human Investigation Committee at the University of Virginia.
The median age (interquartile range [IQR]) for patients with low-stage fibrosis was significantly lower than that for patients with high-stage fibrosis: 46 years (IQR, 38–51) versus 59 years (IQR, 52–64), P = .003. The median BMIs of low-stage and high-stage groups were not significantly different: 36 (IQR, 32–40) versus 37 (IQR, 31–41), P = .91. More female patients were present in both groups, but no difference was seen in the distribution of gender between groups: 79% female versus 71% female, P = .70.
In each stage group, there was no statistically significant difference in the quantity of the MMC among zones. Without separation by fibrosis stage, the abundance of the MMC was not different between zone 1 and zone 3 and between the combined zone (1 and 2) and zone 3 (P = .23 and .26, respectively). Additionally, the abundance of the MMC for each zone was not different from that of an identical zone of different stage group. In the low-stage group, the median scores for the abundance of the MMC in zone 1, zone 2, and zone 3 were 2.30 (IQR, 0.15–2.85), 1.70 (IQR, 0.70–2.70), and 1.15 (IQR, 0.75–2.08), respectively, and in the high-stage group, the median scores for the abundance of the MMC in zone 1, zone 2, and zone 3 were 2.40 (IQR, 0.35–2.73), 1.50 (IQR, 0.28–3.00), and 1.25 (IQR, 0.0–2.50), respectively (Fig. 3). Combining the scores for all three zones also yielded no significant difference in the abundance of MMC between low- and high-stage groups (P = .56).
Similar results were obtained if liver specimens from patients who were taking antioxidative, antidiabetic, and antihyperlipidemic agents were excluded from data analysis. Additionally, a statistically significant difference in the abundance of the MMC from the liver specimens of patients who were taking antioxidative, antidiabetic, and antihyperlipidemic agents and those who were not taking these medications could not be detected by the Wilcoxon rank-sum test. Fewer MMC were identified in the specimens obtained from two cryptogenic cirrhotic patients (patients 9 and 10, Table 2), but the small sample size in this group precluded statistical testing.
Immunoreactivity for 4-HNE or the abundance of the ballooned hepatocytes was not statistically different when compared between low- and high-stage groups (P = .88 and .42, respectively). The abundance of MMC was not significantly different when compared between groups with low (1–2) and high (3–4) immunoreactivity for 4-HNE and between groups with low (0–1) and high (2–3) ballooned hepatocyte score (P = .78 and .54, respectively).
It is now generally agreed that the presence of MMC represents an abnormal condition in various liver diseases. The zonal distribution of these structures has not previously been examined closely, and the function of these structures within the hepatocytes remains undetermined in NASH. Unlike previous findings in alcohol-related liver disease,4 our results show that there is no difference in the abundance of the MMC between low-stage NASH (fibrosis stages 1 and 2) and high-stage NASH (fibrosis stages 3 and 4). These structures seem to be randomly distributed among zones, regardless of the fibrosis stage of NASH (Fig. 3). Their abundance does not correlate with the degree of oxidative stress or abundance of ballooned hepatocytes.
Collective data from previous studies indicate that the formation of the MMC reflect hepatic cellular injury on the biochemical level that can be identified by the impairment of mitochondrial electron transport chain or alteration in oxidative process. For example, Krahenbuhl13 showed that altered mitochondrial morphological features were mirrored by the impairment of the mitochondrial electron transport chain, adenosine triphosphate synthesis, or both. Similar findings were demonstrated in 38 NASH patients by Perez-Carrera et al.14 In the state of increased oxidative stress, more free radicals are formed and thus may inflict damage to the mitochondrial lipid membranes. Sanyal et al.,6 noted greater staining of 3-nitrotyrosine, a marker of oxidative stress and significantly more abundant MMC in NASH, compared with simple steatosis. Our results could not detect a correlation between the abundance of the MMC and the degree of oxidative stress determined by staining for 4-HNE. A difference in the degree of oxidative stress between low- and high-stage NASH also was not identified. A similar pattern was found when hepatocellular injury as represented by ballooned hepatocytes was compared. It is important to note that all of our patients had NASH and none had simple steatosis or steatosis with only inflammation. Thus, a direct comparison between our findings and those of Sanyal et al. cannot be made from this study.
It remains, however, that the presence of the MMC may represent an adaptive process to oxidative stress rather than a secondary injury. It has been shown that lipid metabolism or oxidation, as reflected by β-hydroxybutyrate level and enzyme activity of β-hydroxybutyrate dehydrogenase, are actually higher in both NASH and alcoholic patients, suggesting that the organelle has undergone an adaptive alteration.6, 15 If the MMC represent cellular injury, it is logical to believe that they would be more abundant in zone 3 because of the concentration of injury in this region as observed by light microscopy10 and that there would be a positive correlation between the abundance of the MMC and that of ballooned hepatocytes. This absence of expected zonal distribution along with the lack of an association with ballooned hepatocytes does not support the formation of MMC as a direct result of cellular injury, at least as determined by light microscopic findings. Similar adaptive changes in mitochondrial function have been suggested from work with uncoupling protein in animal models of steatosis.16
Crystalline inclusions have been described in a variety of other conditions. Notably, the findings of similar crystal-like structures in Escherichia coli under conditions of oxidative stress are of interest. Wolf et al.17 demonstrated that these structures were formed from copolymerization of a ferritin-like compound and bacterial DNA. It was hypothesized that the crystalline configuration provides a measure of protection to the bacterial DNA. Because of the evolutionary origin of mitochondria from symbiotic prokaryotes and the important role of iron metabolites in oxidative stress, it is interesting to speculate that the MMC in NASH serve some similar function in an altered metabolic milieu of steatosis.
Our results differ in some respects from past studies. According to Petersen,18 more abundant MMC were observed within zone 1 regardless of the degree of steatosis in alcoholic, diabetic, and obese patients, although the fibrosis stage was not noted. In addition, Junge et al.19 reported that more elongated megamitochondria were localized in zone 1, whereas round to oval megamitochondria were more abundant in zone 3. The explanation for these discrepancies is not certain, although a possibility is that we did not differentiate the various shapes of the megamitochondria in the present study. Our results also differed from the work of Chedid et al.,4 in which more megamitochondria were found in milder forms of alcoholic steatohepatitis. However, Chedid et al. suggested that this represented an adaptive role of the megamitochondria in an early and more benign disease stage. If so, the adaptive process seems to be more random and diffuse in NASH compared with alcohol-related injury.
The significant age difference between the low-stage versus high-stage NASH patients in our study is consistent with a number of past studies showing age as a risk factor for more severe disease.20–23 Additionally, fewer MMC identified in the liver specimens of patients with cryptogenic cirrhosis (patients 9 and 10, Table 2) concur with our prior result.5 This finding supports the absence of the MMC when the disease is burned out and their presence in association with steatosis, although the small sample size of this group prevents formal testing.
The relative abundance and zonal distribution of the MMC in both stage groups were not influenced by the use of antioxidative, antidiabetic, and antihyperlipidemic agents. Even so, we cannot rule out the possibility that these medications could have altered the biochemical behavior of the hepatocytes and their mitochondria. The clinical implication of this finding is unclear, and whether these medications are effective in treating NASH will depend on future clinical trials.
As with any liver disease, sampling error in liver biopsy could be a limiting factor. According to the study by Bedossa et al.24 in hepatitis C patients, the length of biopsy specimen of at least 25 mm is necessary to evaluate fibrosis accurately. Although this study involved only hepatitis C patients, inadequate sample size could be expected to be important in NASH. In our study, each core liver biopsy specimen obtained by the standard percutaneous technique was judged to be an adequate representation of a patient's liver histological features, based on the sample size (core length, 2.5–3.5 cm) and the number of portal tracts observed. To reduce sampling error further at the ultrastructural level, two nonadjacent area were cut from each core specimen that were thin-sectioned further to several other layers.
In conclusion, we found that there is no difference in the abundance of MMC between low-stage (fibrosis stages 1 and 2) and high-stage (fibrosis stages 3 and 4) NASH. These abnormal mitochondria seem to be randomly distributed among zones, regardless of the fibrosis stage. The exact function and biochemical nature of these structures remain to be defined. Histologically, their presence does not seem to correlate with cellular injury as represented by ballooned hepatocytes, and on the biochemical level, there seems to be lack of an association between the abundance of the MMC and degree of oxidative stress as defined by staining for 4-HNE.
The authors thank Anita Impagliazzo Hylton and John Harms for their expertise and assistance in preparing this manuscript and the figure.