Hepatic Hyperplasia in Noncirrhotic Fatty Livers: Is Obesity-related Hepatic Steatosis a Premalignant Condition?

Obesity is associated with an increased risk for cancer of the liver and other epithelial organs including the uterus, colon, and pancreas (1, 2, 3). The mechanisms responsible for this association are unknown. Cirrhosis is generally considered to be most important risk factor for HCC,² because almost all HCCs are diagnosed after cirrhosis has developed, regardless of the original cause of liver damage (4, 5). Obesity is associated with an increased risk of cirrhosis (6, 7), and this may explain why the prevalence of liver cancer is increased in obesity. On the other hand, a growing body of epidemiological evidence from large, population-based studies is beginning to implicate obesity-related hyperinsulinemia and insulin resistance in the pathogenesis of other obesity-related malignancies, such as nonhereditary colon cancer and postmenopausal breast cancer (1, 2, 3, 8, 9, 10).

During the neoplastic process, epithelial hyperplasia and dysplasia generally precede cancer by many years. For example, systematic analyses of colonic or breast epithelia during the various stages of neoplasia demonstrate the progressive accumulation of defects that increase cellular survival, disrupting normal tissue homeostatic mechanisms that balance apoptosis and proliferation (11). A similar process occurs during virally mediated or toxin-related hepatic carcinogenesis (12, 13). Hence, it is reasonable to suspect that hepatic hyperplasia (i.e., increased hepatocyte proliferation relative to apoptosis) also antedates the appearance of liver cancer during insulin-resistant states, such as obesity. However, to our knowledge, there is no information about whether or not hepatocyte hyperplasia occurs during obesity per se. If it does, then it is important to learn whether this potentially premalignant condition begins before the liver becomes fibrotic, because obesity-related hepatic steatosis (fatty liver) is very common, occurring in at least 20% of American adults (14), whereas, only a small minority of these individuals will ever develop significant hepatic fibrosis or cirrhosis (15).

To determine whether, and when, hepatic hyperplasia might occur during obesity, we assessed hepatocyte proliferative and apoptotic activities in the livers of genetically obese, ob/ob mice. ob/ob mice have a naturally occurring mutation in the ob gene that prevents the synthesis of leptin, an adipocyte hormone that regulates satiety and energy balance (16, 17). Consequently, these mice become obese at an early age. Similar to many other murine models of obesity, including those that result from diabetes (db) and fatty (fa) loss-of-function mutations in the ob (leptin) receptor, young ob/ob mice develop insulin resistance, hyperinsulinemia, and eventually become hyperglycemic (i.e., type 2 diabetic; Refs. 18, 19, 20). An increased incidence of HCC has been demonstrated in old mice that are homozygous for the ob mutation, even when this defect is present in C57BL-6J strains that are generally resistant to malignancy (21). Interestingly, from a very young age, ob/ob C57BL-6J mice have enlarged, fatty livers (22). Although such mice are unusually vulnerable to toxic and ischemic liver injury and have been used as models for the early stages of human NAFLD (23), they do not develop cirrhosis spontaneously (24). Thus, despite the strong association between cirrhosis and HCC in humans, it is unlikely that cirrhosis accounts for the increased prevalence of HCC in ob/ob mice. Hence, these mice are relatively “pure” models of obesity-related hepatic insulin resistance and liver cancer that can be studied at a premalignant stage to determine whether, and when, hepatic hyperplasia develops. Our results demonstrate that, in the absence of any overt, exogenous insult, young adult male ob/ob mice have increased hepatocyte proliferation relative to apoptosis and suggest that this imbalance promotes increased liver mass. Given the general temporal relationship between epithelial hyperplasia and eventual neoplasia, these data support the possibility that cellular survival advantages develop in the liver during insulin-resistant states and contribute to obesity-related hepatic neoplasia.

Eight-week-old ob/ob C57BL-6J mice (n = 12) and their lean (heterozygote) litter mates (n = 12) were purchased from Jackson laboratories (Bar Harbor, ME). The mice were maintained in a temperature- and light-controlled animal facility and permitted ad libitum access to water and standard pellet-type chow for 2 weeks. At the end of this period, one mouse was randomly selected from each group, anesthetized with phenobarbital and evaluated by ¹H-NMR-spectroscopy to quantify fat accumulation in the liver (25). Subsequently, all of the mice were weighed, given injections of BrdUrd i.p., and killed 2 h later by cervical dislocation. The livers were removed from the carcasses and weighed, and then a small biopsy was obtained from each organ. This tissue was placed in buffered formalin for subsequent evaluation of hepatic histology and nuclear BrdUrd incorporation. The remaining liver tissues were used immediately to isolate either fresh nuclei (n = 4 mice/group) or fresh mitochondria (n = 4 mice/group). Others (n = 4 mice/group) were snap-frozen in liquid nitrogen, stored at −80°C and processed subsequently to obtain whole liver extracts of RNA, DNA, and protein.

This was assessed by counting BrdUrd (+) hepatocyte nuclei on coded liver sections that were counterstained with hematoxylin. Two sections from each mouse were evaluated independently by two observers (S. Y., J. H.). In each section, 10 ×400 were examined and the total number of hepatocytes with and without BrdUrd (+) nuclei were recorded. An average proliferative index (number of BrdUrd (+) hepatocytes ÷ total number of hepatocytes) was calculated for each mouse; then the figures from the 12 mice in each group were totaled to obtain the mean (± SE) proliferative index for each group. In addition, freshly isolated liver cell nuclei from four mice/group were stained with propidium iodide, and nuclear ploidy was evaluated by flow cytometry.

Members of the MAPK family, Erk-1 and Erk-2, become phosphorylated (i.e., activated) during the prereplicative phase of hepatocyte proliferation. Hence, steady-state levels of phospho-Erk-1 and -Erk-2 are useful measures of hepatocyte proliferative activity. The content of total and phosphorylated Erk-1 and Erk-2 were evaluated in whole liver homogenates from four mice/group using immunoblot analysis with specific antisera from Santa Cruz Biotechnology, Inc (Santa Cruz, CA) as described previously (26).

Total hepatic RNA was isolated from the livers of four mice/group using a minor modification (27) of Chomcznski and Sacchi’s method (28). RNase protection analysis of this RNA was performed using reagents from PharMingen as we have described previously (29). This assay permits the concurrent evaluation of several immediate early genes, such as c-jun, delayed-early genes, such as cyclin-D1, and S-phase-associated genes, such as cyclin A, as well as housekeeping genes, such as GAPDH, in the same RNA sample, facilitating standardization that controls for slight variability in RNA content that might occur among different samples.

Hepatic DNA was extracted and quantitated as described by Blin and Stafford (30). Results are expressed as mg/g liver and also as mg/total liver.

The optimal technique for assessing apoptosis in the liver remains hotly contested. Standard approaches, including TUNEL staining and evaluation of DNA fragmentation by agarose gel electrophoresis, have been criticized because these assays may also become positive in necrotic livers (31). Detection of apoptotic hepatocytes on H&E-stained tissue sections is generally acknowledged to be reliable. Thus, we used the latter approach to screen for differences in basal apoptotic rates between ob/ob and lean mice. Coded sections were reviewed by an experienced hepatologist (A. M. D.) who counted the numbers of apoptotic hepatocytes (characterized as shrunken, hypereosinophilic cells with pyknotic or absent nuclei) in 10 randomly selected ×200 microscope fields on each of two sections/mouse. The mean (± SE) number of apoptotic hepatocytes/10 microscope fields was calculated from 12 mice in each group. However, because this approach risks underestimating apoptosis in the severely fatty livers of ob/ob mice, alternative measures of apoptosis were also assessed.

Whole liver homogenates from four mice/group were evaluated for evidence of caspase 3 activation. Equal amounts (40 μg/lane) of liver protein from ob/ob and lean mice were separated by SDS-PAGE, transferred to nylon membranes and probed for total caspase 3 and its cleavage products using antisera from PharMingen (San Diego, CA). The cleavage of procaspase 3 to caspase 3 activates the enzyme during hepatocyte apoptosis. Moreover, caspase 3 is the major effector of apoptosis in hepatocytes. Thus, the ratio of caspase 3 cleavage products (i.e., active caspase 3):procaspase 3 is a good measure of hepatocyte apoptotic activity (32). The contents of pro- and active-caspase 3 were determined in each sample of liver protein by densitometric analysis of the immunoblots and were expressed as a ratio of activated caspase 3:total caspase 3 in each sample. Mean (±SE) results were derived from 4 mice/group.

The basal rate of apoptosis is extremely low in normal adult livers (33), challenging the sensitivity and reliability of assays that attempt to detect and quantitate further reductions in apoptotic activity (31). However, if basal apoptotic activity is truly inhibited, then it is reasonable to assume that antiapoptotic mechanisms might be induced. The latter include increased DNA binding activity of NFκB (34), accumulation of bfl-1 and bcl-xL, antiapoptotic bcl-2 family members (35), and activated antioxidant defenses, such as mitochondrial MnSOD (36). According to previously published methods (27), nuclear protein extracts (8 μg of protein/assay) from four mice/group were analyzed individually by EMSAs using ³²P-labeled, double-stranded oligonucleotide fragments that contain the κB binding site in the immunoglobulin enhancer element (37). In some assays, 1 μl of antisera specific for NFκB p65 or p50 subunits was added to verify that the complexes contained NFκB. In four other mice/group, the activity of a NFκB target gene, MnSOD (34), was evaluated in hepatic mitochondrial extracts and normalized per microgram of total mitochondrial protein as described previously (38). Finally, RNase protection analysis was done using a 10-μg sample of total liver RNA from the final four mice in each group (29). Reagents from PharMingen were used to assess the expression of bfl-1, bcl-xL, and GAPDH in each sample.

ob/ob mice have hepatomegaly. Because ob/ob mice weigh about twice as much as their lean litter mates, it is not unexpected that their livers are larger than those of lean controls. However, normalization of liver weight by body weight demonstrates that the mass of ob/ob livers is greater than predicted solely on the basis of generalized obesity (Table 1,). Given that ob/ob mice have fatty livers (Fig. 1a), it is conceivable that the tremendous accumulation of lipid within individual hepatocytes increases the liver:body weight ratio. However, analysis of the liver and entire body of a typical ob/ob mouse by ¹H-NMR spectroscopy demonstrates that there is actually less accumulation of fat in the liver (about 15% of the total ¹H signal) than there is in the rest of the body (in which fat accounts for almost 60% of the ¹H signal); therefore, this explanation seems unlikely (Fig. 1,b). In ob/ob mice of this age, there is no overt histological evidence for the hepatic accumulation of inflammatory cells or fibrous tissue (Fig. 1,a). Hence, a comparison of the total DNA content of fatty and normal livers should provide information about hepatocyte number and should help to determine whether increases in hepatocyte number contribute to ob/ob hepatomegaly. Consistent with evidence that ob/ob hepatocytes contain more fat than normal hepatocytes, the hepatic DNA content normalized per gram of liver wet weight is about 50% lower than normal. However, ob/ob livers weigh almost four times more than normal. Hence, the total DNA content of an ob/ob liver is, on average, significantly greater than that of an age- and gender-matched lean litter mate (Table 1 ), which suggests that increases in hepatocyte number contribute to the increased liver mass of fatty livers.

In adults, liver mass is maintained at a relatively constant value because the rates of hepatocyte proliferation and apoptosis are similar (39, 40). Moreover, unlike cells in other epithelial tissues, hepatocytes “turn over” infrequently in uninjured livers, as evidenced by observations that at any given moment, very few hepatocytes incorporate [³H]thymidine or BrdUrd into nuclear DNA (33). When the normal balance between proliferation and apoptosis is disturbed to favor a net increase in proliferative activity, liver size increases. This can occur either as a result of factors that work predominately by increasing hepatocyte proliferation or as a consequence of factors that work predominately by decreasing hepatocyte apoptosis. The largest increases in liver mass occur when hepatocyte proliferation and apoptosis are regulated differentially (i.e., when proliferation is stimulated while apoptosis is inhibited). Although tissue neoplasia typically involves increased proliferative activity, it is not clear that the latter process contributes to the hepatomegaly of ob/ob livers. Indeed, the possibility that basal rates of hepatocyte proliferation might be decreased in ob/ob livers merits consideration because it was recently reported that liver regeneration is inhibited after partial hepatectomy in obese, diabetic, fa/fa rats with fatty livers (41).

To compare the basal proliferative activities of normal and ob/ob livers, several approaches were used. First, we looked for evidence of activation of the MAPK cascade in the two groups. The content of phosphorylated MAPK isoforms (p42 Erk-1 and p44 Erk-2) is more than 4-fold greater in fatty, ob/ob livers than in normal livers despite similar content of total Erk-1 and Erk-2 in the two groups (Fig. 2,). MAPK activation typically occurs within minutes after hepatocytes are exposed to growth factors (33). Another early event that heralds the reentry of hepatocytes into proliferative phases of the cell cycle is the accumulation of mRNAs that encode immediate-early genes, such as c-jun (42, 43). RNase protection analysis of liver RNA from normal and fatty livers demonstrates that c-jun mRNAs are increased in fatty livers (Fig. 3,). Moreover, transcripts for cyclin D₁ and cyclin A, genes that are up-regulated at the time of transition from G₁ into S-phase (44, 45), are also increased in fatty livers (Fig. 3,). The induction of cell cycle-associated genes in fatty livers suggests that hepatocyte DNA synthesis might be increased in ob/ob mice. To evaluate that possibility more directly, liver nuclei were isolated; stained with propidium iodide, and evaluated for DNA content by flow cytometry (46). An increased proportion of nuclei with hyperdiploid DNA content were demonstrated in fatty livers (Fig. 4; Table 2,). Consistent with these results, BrdUrd incorporation into hepatocyte nuclei is increased in tissue sections prepared from fatty livers (Table 2 ).

Although hepatocyte proliferative activity is increased slightly in ob/ob livers, this will not increase liver mass if the rate of hepatocyte apoptosis increases proportionally (39, 40). During sublethal exposure to various noxious agents, hepatocyte apoptosis increases and liver mass is preserved by an associated increase in hepatocyte proliferative activity (33). Because the accumulation of fat within hepatocytes (hepatic steatosis) occurs in the early stages of liver disease that occurs during obesity (47), it is conceivable that the increased proliferative activity of ob/ob hepatocytes is merely an adaptive response to increased hepatocyte apoptosis. To evaluate this possibility, we counted the numbers of apoptotic hepatocytes in H&E-stained liver sections from four ob/ob mice and an equal number of their lean litter mates. To our surprise, we were unable to demonstrate an increased accumulation of apoptotic hepatocytes in ob/ob livers. In fact, an inspection of 10 ×200 microscope fields/liver section demonstrated an average of three to four apoptotic foci/section in control mice but fewer than two apoptotic foci/section in ob/ob livers, which suggests that ob/ob livers might actually have fewer apoptotic hepatocytes than do normal livers. However, because the basal rate of hepatocyte apoptosis in adults is very low, it is difficult to detect a further reduction in the already-low number of apoptotic hepatocytes simply by inspecting tissue sections. In addition, the accumulation of fat in ob/ob hepatocytes might have obscured a small, but significant, increase in apoptotic cells. To overcome these technical obstacles, alternative measures of apoptotic activity were evaluated.

The cleavage of procaspase 3 activates this enzyme, which functions as the predominant terminal effector of hepatocyte apoptosis (48). Thus, the ratio of procaspase 3 cleavage products (i.e., activated caspase 3) to total caspase 3 content provides a measure of apoptotic activity (32). To compare the degree of caspase 3 activation in ob/ob and normal livers, equal amounts of whole liver protein from fatty and lean livers were separated by SDS-PAGE and exposed to commercially available antisera that recognize both procaspase 3 and its cleavage products. Consistent with the histological evidence for decreased apoptosis in ob/ob livers, the immunoblot assays demonstrated a significantly reduced ratio of activated:total caspase 3 in livers from ob/ob mice compared with livers from their lean litter mates (Fig. 5,a). To validate this approach, a separate group of lean C57BL-6 mice were given injections i.p. of a small dose (10 μg/mouse) of bacterial LPS, which is known to cause a slight increase in hepatocyte apoptosis (49). Whole liver homogenates were analyzed for procaspase 3 and its cleavage products by immunoblot (Fig. 5 b). As expected, LPS treatment increases the ratio of activated caspase 3:total caspase 3 in normal mice, which confirms that this assay is a useful measure of hepatocyte apoptosis.

Taken together, the results suggest that basal rates of hepatocyte apoptosis are actually decreased in young adult ob/ob mice. If this suspicion is correct, then one or more of the cellular mechanisms that inhibit apoptosis should be induced in ob/ob livers. The transcription factor, NFκB plays a critical role in hepatocyte antiapoptotic defense (50). A comparison of NFκB-DNA binding activity in the ob/ob mice with fatty livers with that of age- and gender-matched normal mice of the same strain demonstrates increased NFκB-DNA binding activity in nuclear extracts from fatty livers (Fig. 6,). The production of superoxide anion (O₂⁻) by mitochondria increases even before the induction of NFκB early during the course of apoptosis (51), and it has been shown that apoptosis is inhibited in cells that are transfected with mitochondrial MnSOD, the enzyme that detoxifies O₂⁻ (36, 52). To clarify events that might protect ob/ob hepatocytes from apoptosis, mitochondrial MnSOD activity was compared in ob/ob mice and their lean litter mates (38). Hepatic mitochondrial MnSOD activity is greater in ob/ob mice (Fig. 7,). Activation of mitochondrial MnSOD is predicted to decrease oxidative stress (36), and, hence, this finding is difficult to reconcile with the EMSA data that demonstrate induction of NFκB, a redox-sensitive transcription factor (53), in ob/ob liver nuclei. However, overexpression of certain antiapoptotic members of the Bcl-2 family of mitochondrial membrane proteins, such as Bfl-1 or Bcl-xL, can inhibit apoptosis after cytochrome c release and NFκB activation (35, 54, 55, 56). Fatty livers from ob/ob mice have accumulated an increased level of bfl-1 and bcl-xL transcripts (Fig. 8 ). Thus, several antiapoptotic factors, including NFκB, MnSOD, bfl-1, and bcl-xL, have been induced in ob/ob livers, providing potential explanation(s) for the reduced activation of caspase 3 and decreased accumulation of apoptotic hepatocytes in ob/ob mice. Although the absolute magnitude of the difference between ob/ob and lean livers for any one of these antiapoptotic responses is relatively small, the combination of inhibited apoptosis plus slight increases in hepatocyte proliferation cause hepatic epithelial hyperplasia and contribute to the increased mass of ob/ob livers.

To our knowledge, these studies are the first to demonstrate an association between obesity and hepatocyte hyperplasia. Because the work was done with genetically obese mice that were housed in a tightly monitored animal facility, it is unlikely that the findings can be explained by occult exposure to environmental toxins or hepatotrophic viruses. Moreover, the observed increases in hepatocyte proliferation and decreases in hepatocyte apoptosis occurred in the absence of overt hepatic inflammation or fibrosis and did not require exposure to tumor promoting drugs. In this regard, the fatty livers of young, adult ob/ob mice resemble the livers of myc- or TGFα-transgenic mice and myc/TGFα-double transgenic mice, in which increased liver mass is induced experimentally by the overexpression of a known proto-oncogene and/or hepatocyte growth factor (57, 58). Similar to these transgenic mice (58, 59), aged ob/ob mice also have an increased incidence of HCC (21, 60, 61), although the tumor incidence rate appears to be much greater in the former. Nevertheless, given the acknowledged association between obesity and cancer in several organs, including the liver (1, 8, 9), evidence for obesity-related increases in hepatocyte proliferative activity relative to apoptosis have direct clinical relevance.

Although the mechanisms for the association between obesity and liver cancer are unknown, cellular survival advantages that occur during insulin resistance may contribute to hepatocarcinogenesis because obesity-related insulin resistance correlates with an increased risk for cancers in other organs, such as the breast and colon (2, 3). In general, most cases of HCC are diagnosed in patients who have had cirrhosis for many years (4, 5). However, it is not clear whether the neoplastic process begins during cirrhosis or starts at earlier stages of liver disease. If HCC begins during cirrhosis, then it is reasonable to speculate that hepatic insulin resistance might be involved, because advanced liver disease is generally associated with insulin resistance, and the association between cirrhosis and diabetes is particularly strong in obese individuals (7). On the other hand, it has been difficult to determine whether insulin resistance plays an independent role in the pathogenesis of obesity-related hepatic neoplasia or is merely an associated abnormality that, like HCC, develops during cirrhosis.

Obesity is strongly associated with a spectrum of liver diseases that have been dubbed NAFLDs because they resemble alcohol-induced fatty liver diseases (6, 62, 63). Fatty liver (hepatic steatosis) is the earliest and most common stage of NAFLD, afflicting almost 20% of American adults (14), whereas cirrhosis, the most advanced stage of NAFLD, develops over decades in only a minority of these individuals (15). Some degree of insulin resistance is present in most NAFLD patients with hepatic steatosis (64), and, as with other types of liver disease, the severity of hyperinsulinemia and insulin resistance increase as cirrhosis becomes established (64, 65, 66). However, because insulin resistance clearly antedates cirrhosis in many patients with NAFLD, livers could be evaluated during the early stages of NAFLD to determine whether hepatocyte hyperplasia occurs during insulin resistance when the liver is not cirrhotic. Genetically obese ob/ob mice with fatty livers have many of the features of the insulin-resistance syndrome in humans (18, 22). Hence, the livers of these mice can be analyzed to determine whether the process of hepatocyte neoplasia begins during the early stages of obesity-related fatty liver disease, when there is insulin resistance but not cirrhosis.

The present results demonstrate clearly that hepatocyte hyperplasia occurs at the fatty liver stage of NAFLD. However, as will be discussed subsequently, whether or not insulin resistance has any direct role in this process remains uncertain. The fatty hepatocytes of ob/ob mice exhibit many similarities to liver cells that have been transformed experimentally with the ras or raf proto-oncogenes. For example, both fatty livers and ras transformation result in MAPK activation, NFκB induction, and the up-regulation of antiapoptotic bcl-2-related genes (67), that are transcriptional targets for NFκB (68). Others have demonstrated that NFκB is a downstream target of the MAPK cascade (69) and have shown that NFκB induction required for raf- or ras-neoplastic transformation of rat liver epithelial cells (67), which suggests that the induction of NFκB and MAPK might contribute to the hepatocyte hyperplasia and hepatomegaly that occur in obesity-related fatty livers. However, if so, then it is paradoxical that hepatic NFκB induction occurs in insulin-resistant ob/ob mice, because there is compelling evidence that insulin itself activates NFκB through a Raf-1-mediated pathway in other mammalian cells (70). Indeed, the ability of insulin to induce NFκB is required for the antiapoptotic actions of the hormone in several cell culture models (71).

Hence, it is conceivable that, rather than insulin resistance per se, one of the other hormonal, immunological, and metabolic abnormalities in obese, ob/ob mice might mediate hepatocyte hyperplasia. Given the association between type-2 diabetes, obesity, and malignancy in humans (8, 9) and experimental animals (21, 60, 61), the role of hyperinsulinemia (as opposed to insulin resistance) in the neoplastic process certainly merits investigation. In light of the recent demonstration that reactive oxygen species are required for hepatocarcinogenesis in c-myc/TGFα transgenic mice (72), the possibility that obesity-related increases in hepatic oxidative stress contribute to hepatic hyperplasia should also be considered. Hepatic mitochondrial production of reactive oxygen species is increased significantly in ob/ob mice (38). Thus, the possibility that oxidant stress might be one of the mechanisms driving hepatic hyperplasia in these animals is a particularly intriguing hypothesis. Although the mechanisms driving increased reactive oxygen species release from ob/ob liver mitochondria are unknown, leptin deficiency and secondary changes in cytokine production (29, 73, 74) might be involved. If the latter is true, then additional work is necessary before data from our studies with leptin-deficient ob/ob mice can be extrapolated to obese humans with fatty livers, because hyperleptinemia (rather than leptin deficiency) is the rule in human obesity (75).

Despite leaving many questions unanswered, our work is important because it demonstrates that some of the mechanisms that promote HCC in NAFLD, one of the most common types of liver disease, become operative “spontaneously,” i.e., without an obvious requirement for comorbidity factors, such as viral infection or hepatotoxin exposure. Moreover, this hyperproliferative response begins long before cirrhosis occurs. Indeed, given evidence that ob/ob mice are resistant to cirrhosis (24), but eventually develop HCC (21), our observations raise the intriguing possibility that obesity-related fatty liver, itself, is a premalignant condition. If this possibility is confirmed by others, then our recent findings extend other data (7, 47, 76) that indicate that NAFLD is not always a benign process.