Hepatocellular neoplasms after intrahepatic transplantation of ovarian fragments into ovariectomized rats


  • Potential conflict of interest: Nothing to report.


Intrahepatic transplantation of ovarian fragments in ovariectomized rats results in morphological abnormalities. The liver acini draining blood from ovarian grafts show alterations resembling chemically induced amphophilic hepatocellular preneoplasias. We investigated the long-term development of these estrogen-induced foci of altered hepatocytes (FAH). We divided 451 Lewis rats into one main group (MG) and 11 (7 female, 4 male) control groups and observed them for up to 30 months. MG animals were ovariectomized and received ovarian transplants into the right liver part. Different combinations of castration, transplantation of ovarian or testicular fragments, and administration of antiestrogenic toremifene were used in controls. In the MG, transplants showed signs of gonadotropic stimulation, and estrogen levels were strongly increased in the downstream liver acini. After 6 and 12 months, FAH developed in hepatocytes downstream of the transplants. After 18 months, 27% of the MG animals showed transformation of FAH into hepatocellular adenomas; this figure increased to 42% after 24 months (8/19), significantly outnumbering four spontaneous adenomas that developed between 18 and 30 months in 258 control animals. Hepatocellular carcinoma (HCC) appeared only in the MG. At 24 and 30 months, 18 HCCs developed; thus, 78% of MG animals showed at least one carcinoma. Administration of toremifene in ovariectomized and transplanted animals completely prevented hepatocarcinogenesis. Testicular grafts showed no influence on liver tissue. In conclusion, initially adaptive but preneoplastic alterations in hepatocytes downstream of intrahepatically transplanted ovarian fragments may transform into HCC, indicating a strong hepatocarcinogenic potential of high local levels of endogenous estrogens in the rat liver. (HEPATOLOGY 2006;43:857–867.)

Oral contraceptive steroids contribute to the development of hepatocellular adenoma (HCA) in humans, and a recent meta-analysis has also shown an increased incidence of hepatocellular carcinoma (HCC) in women of a low-risk population (low hepatitis B and C prevalence).1 These observations correspond to findings in animal experiments that have demonstrated an enhancement of hepatocarcinogenesis after administration of various synthetic estrogens such as ethinyl estradiol and diethylstilbestrol.2–6 Simultaneous administration of tamoxifen inhibited part of the carcinogenic effects of ethinyl estradiol, indicating that they were most likely mediated by the estrogen receptor.7 These observations suggest that estrogens are not only genotoxic8, 9 but may also act as hepatocarcinogens by altering cell signaling and metabolism, including nuclear and mitochondrial gene expression and cell proliferation.10, 11 Estrogens also interact with other hepatocellular mitogens such as epidermal or hepatocyte growth factor and transforming growth factor α (TGF-α).12–14 Although the carcinogenic effects of synthetic estrogens on the liver have been extensively studied, little is known about the hepatocarcinogenic potential of endogenous estrogens, although increased levels of circulating estrogens bear a risk for the development of certain human malignant neoplasms, such as breast and endometrial cancer.15, 16 We have previously shown that high local levels of endogenous estrogens, derived from hormonally stimulated ovarian fragments after their intrahepatic transplantation in ovariectomized rats, exert characteristic focal alterations in the downstream liver parenchyma.17 The changes in these foci of altered hepatocytes (FAH) included loss of glycogen, cytoplasmic amphophilia, nuclear enlargement, a strong increase in peroxisome density, and an increase in proliferative activity, as well as apoptotic elimination and changes in the activities of certain key enzymes of the energy metabolism. The phenotype of these FAH resembled in many respects that of amphophilic preneoplastic foci that appear after long-term administration of different chemical hepatocarcinogens.18–20 Preliminary data also indicated that some of these lesions progress to hepatocellular neoplasms.17 However, dietary administration of the natural estrogen estradiol did not show any hepatocarcinogenic effect in contrast to similarly administered synthetic estrogens in another study.21 Thus, the carcinogenic potential of endogenous estradiol in the liver still needs to be clarified. Moreover, ample experimental and epidemiological data obtained from animal and human studies indicate that synthetic androgens, mainly anabolic steroids, are also tumorigenic in the liver,22–26 and HCC is generally more common in males than in females, indicating a possible tumorigenic influence of endogenous testosterones.27

In this long-term study, we therefore examined the development of the putative preneoplastic FAH after transplantation of ovarian fragments into female rats, tested the effect of concurrently administered antiestrogenic toremifene, and also investigated the effect of natural testosterone in male rats, using a similar experimental setting with castration and intrahepatic gonadal (testicular) tissue transplantation.


HCA, hepatocellular adenoma; HCC, hepatocellular carcinoma; TGF-α, transforming growth factor α; FAH, foci of altered hepatocytes; MG, main group; CG, control group; PAS, periodic acid-Schiff.

Materials and Methods


Female and male 3-month-old highly inbred Lewis rats (n = 451) weighing approximately 200 g (Harlan & Winkelmann, Borchen, Germany) were used in this study. They were housed as previously described.17 All experiments were performed in line with the guidelines of the Society for Laboratory Animals Service (GV-Solas).

Experimental Groups.

The animals were subdivided into one main group (MG) and eleven control groups (CGs). They were treated as shown in Table 1.

Table 1. Experimental Design and Number of Hepatocellular Tumors
 MGCG 1CG 2CG 3CG 4CG 5CG 6CG 7CG 8CG 9CG 10CG 11
  • NOTE. After 6 and 12 months, no tumors were observed (not listed). All tumors in the MG originated from the right part of the liver. Four sporadic adenomas developed in old animals of the CGs. The number of investigated animals is given in parentheses.

  • Abbreviations: MG, main group, CG, control group; F, female; M, male.

  • *

    Significantly more tumor-bearing animals than in each CG 1– 5.

  • Significantly more carcinoma-bearing animals than in each CG 1–5.

  • After 24 months, 14 animals had developed HCC, and 5 showed additional HCA. Three animals exhibited only HCA.

Number of animals565245514951331632242418
Castration× × × × × × 
Transplantation of ovarian fragments××  ×       
Transplantation of testicular fragments            
Number of hepatocellular adenomas/carcinomas            
 18 months4/0 (15)0 (12)0 (7)0 (12)0 (11)0 (11)0 (3)0 (5)0 (2)0 (3)0 (2)0 (3)
 24 months8/14 (19),*,2/0 (15)0 (9)0 (15)0 (10)0 (11)0 (18)0 (10)0 (16)0 (3)0 (7)0 (5)
 30 months0/4 (4)*,0 (5)1/0 (9)0 (4)1/0 (9)0 (9)0 (6)— (0)0 (6)0 (12)0 (10)0 (8)

Castration and Transplantation.

The gonadal tissue of donor animals was obtained as previously described.17 The ovaries were dissected into small fragments, suspended in ice-cold Hanks' balanced solution (pH 7.2) (Sigma, Heidelberg, Germany), and transplanted via the portal vein into the right part of the liver (i.e., the right lobe, the right part of the middle lobe, and the caudate lobe), secured by previous clamping of the left branch of the portal vein. Each animal received the same amount of gonadal tissue; all ovarian fragments were from one donor animal. The clamp was then removed; thus, ischemia lasted no longer than 1 minute.

Administration of the Estrogen Receptor Antagonist Toremifene.

Two groups, CG 4 and CG 5, received a daily dose (12 mg/kg) of the estrogen receptor antagonist toremifene (Orion Corporation Farmos, Turku, Finland) during the entire experimental period. Toremifene was dissolved in 0.5% carboxylmethylcellulosis (Sigma, Heidelberg, Germany) and administered into the pharynx with a tube. Administration began 2 days before transplantation.

Tissue Preparation.

Animals were killed after 6, 12, 18, 24, and 30 months and were fixed via perfusion as previously described in detail.17 After fixation, the livers were removed and cut into slices. From every liver, 10 specimens of 2 × 2 mm were postfixed with OsO4 and embedded in Epon. Semithin sections were stained with Richardson's stain, and ultrathin sections were stained with uranyl acetate and lead citrate and examined using a Philips CM10 microscope (Philips, Einthoven, The Netherlands). In addition to all macroscopic lesions, 12 slices from each liver (left and right part) were embedded in paraffin and cut into 1- to 2-μm sections. Paraffin sections were also prepared from the heart, lungs, thyroid gland, pancreas, liver, duodenum, kidneys, adrenal glands, spleen, brain, pituitary gland, uterus, mammary gland, and gonads if the animals were not castrated. All sections were stained with hematoxylin-eosin or periodic acid-Schiff (PAS) reaction. Classification of hepatocellular neoplasms was determined according to Metzger et al.18

TGF-α and α-Inhibin Immunohistochemistry.

Paraffin sections (1-2 μm) were pretreated for antigen retrieval after deparaffinization by cooking in a microwave oven in citrate buffer (pH 6.0) for 30 minutes at 750 W. The dilution for the monoclonal anti–TGF-α antibody (clone MF9; Oncogene Science/Dianova, Hamburg, Germany) and α-inhibin antibody (clone R1; Serotec, Oxford, United Kingdom) was 1:10 and 1:5, respectively. Visualization was performed with the LSAB+ kit and DAB+ kit (DAKO, Hamburg, Germany) according to the manufacturer's instructions. Rabbit immunoglobulin G served as a negative control (Oncogene Science/Dianova). Sections were finally counterstained with hematoxylin.

Sex Steroid and Gonadotropin Measurement.

Serum samples from aortal blood were taken at the time of sacrifice and were analyzed for testosterone, 17β-estradiol, progesterone, and luteinizing hormone via radioimmunoassay. An immunoradiometric assay was performed for follicle-stimulating hormone. All assays were purchased from MP Biomedicals Germany GmbH (formerly ICN Biomedicals GmbH, Eschwege, Germany), and determinations were performed in duplicate. The assays for testosterone, estradiol, and progesterone were adapted to smaller samples (25-50 μL), whereas the others were performed according to the manufacturer's original protocol.

Local concentrations of 17β-estradiol in the liver of six additional rats (4 MG animals, one solely ovarectomized rat of CG 2, and one completely untreated rat of CG 3) were determined in serum from blood samples taken both from hepatic veins and directly from the hepatic parenchyma, immediately adjacent to the grafts located in the right liver lobe and from the transplant-free left liver lobe. For this purpose, the 17β-estradiol radioimmunoassay procedure mentioned above was adapted to sample sizes of 5 μL each. Whenever possible, samples were determined in duplicate and the median and standard deviation was determined, which was not possible when the single values were below (<5.3 pg/mL) or above (>2,583 pg/mL) the measuring range.

Statistical Analysis.

Differences regarding the number of tumor-bearing animals were assessed via Fisher exact test. A Wilcoxon-Mann-Whitney test was applied for determining the differences in body weight (supplementary data; Supplementary material can be found on the HEPATOLOGY website: http://interscience.wiley.com/jpages/0270-9139/suppmat/index.html) and in serum hormone levels. A P value of less than .05 was considered statistically significant.


Morphology of the Transplanted Gonadal Tissue

Ovarian Transplants.

Six months after transplantation, white or yellow ovarian transplants were detected stereomicroscopically in the right parts of the livers of all animals receiving ovarian grafts (Figs. 1A-C and 2A,C). Owing to gonadotropic stimulation, the grafts in the MG showed an increase in size when compared with the small fragments that were transplanted 6 months before and exhibited cystic structures (Fig. 1A). Histologically, the grafts consisted of follicles in varying stages of maturation, oocytes, corpora lutea, and corpora albicantia. These different stages of maturation were often observed concurrently and in close spatial relationship within the same graft. Therefore, a “functional status” of the individual transplant could not clearly be determined. The grafts in the animals of CG 1, which were not previously ovariectomized, displayed smaller follicles with considerably less granulosa cells than in the MG. No corpus luteum was found in these animals. The transplants of CG 1 began to become fibrotic 12 months after transplantation and were completely fibrosed at 18, 24, and 30 months. In CG 4, the graft morphology was similar to the MG, though they were slightly smaller (Fig. 2C-D). A small corpus luteum of less than 1 mm was sometimes seen in this group at all time points. Despite similar graft morphology, hepatocellular alterations did not occur in any of the CG 4 animals, which were treated with the estrogen receptor antagonist toremifen. In the livers of the MG, one to five transplants became considerably hyperplastic in all animals after 18 to 30 months, reaching a size of up to 6 mm (Fig. 1F). These hyperplastic transplants did not show follicles with a multilayer of granulosa cells anymore, but consisted of large amounts of α-inhibin–positive stromal cells with variable lipid content and without cytologic atypia, sometimes accompanied by a simple cyst lined with a flat single cell layer.

Figure 1.

Hepatocarcinogenesis in the MG. (A) Stereomicroscopic aspect of an unstained perfusion-fixed liver slice 6 months after transplantation. The ovarian graft measures 2 mm and is composed of solid and cystic structures (i.e., follicles). (B) Two transplants measuring 3 mm were visible 12 months after transplantation. Arrowheads mark preneoplastic FAH, which can be identified as a rim of darker-stained liver tissue adjacent to the ovarian grafts. (C) HCA (5 mm) adjacent to the ovarian graft (bottom). (D) HCC originating from the right part of the liver (ventral view, right liver part at the bottom). (E) HCC including a transplant, which is visible as a yellow focus in the upper right quadrant of the tumor. (F) Two stromal tumors that have originated from ovarian grafts 18 months after transplantation. (G) Histomorphology of FAH 6 months after transplantation. The ovarian graft is visible at the top. The adjacent liver parenchyma (middle part) is characterized by enlarged and thickened hepatocellular trabeculae, while the liver tissue of the neighboring liver acinus, which does not drain the hyperestrogenemic blood from the graft, is unaltered (bottom). (H) Most HCCs were characterized by a trabecular growth pattern with moderate cellular pleomorphism. PAS reaction in this HCC 24 months after transplantation reveals glycogen-storing in some neoplastic cells. Three mitotic figures are visible. (I) Lung metastases of another well-differentiated HCC 30 months after transplantation. The width of each panel represents: (A) 2.1 mm; (B) 6.3 mm; (C) 9.7 mm; (F) 5.6 mm; (G) 375 μm; (H) 240 μm; and (I) 960 μm; (D) and (E) show a macroscopic overview.

Figure 2.

Female control groups. (A) Transplants of CG 1, females that received transplants without prior ovariectomy were small and only rarely showed follicles, as in this animal 12 months after transplantation. The adjacent liver parenchyma showed no morphological alterations. (B-D) Toremifene administration in female rats otherwise treated exactly like the main group completely prevented hepatocarcinogenesis. The livers were morphologically normal (B) (CG 4, 30 months after transplantation), whereas the transplants were vital and showed follicle maturation (C-D) (stereomicroscopic aspect and histomorphology, 18 and 24 months after transplantation, respectively). (E-F) Testicular grafts were vital but did not influence the morphology of the surrounding liver parenchyma (E) (stereomicroscopic aspect 6 months after transplantation) and were composed only of Sertoli and Leydig cells (F) (6 months after transplantation). The width of each panel represents: (A) 5.0 mm; (C) 3.6 mm; (D) 2.9 mm; (E) 14.6 mm; and (F) 700 μm; (B) shows a macroscopic overview.

Testicular Transplants.

Macroscopically, testicular grafts presented as yellowish-white foci in the right part of the livers (Figs. 2E and 3A). Six months after transplantation, the transplants were vital and consisted almost exclusively of Sertoli and Leydig cells (Figs. 2F and 3B-C). The seminiferous tubules became increasingly fibrotic, and germ cells and spermatozoons were almost never detected. Transplants in castrated male and female rats (CG 6 and CG 8) were larger than in the uncastrated animals (CG 7 and CG 9), corresponding to the level of gonadotropin secretion (Table 2). After 30 months, some of the transplants in CG 6 and CG 8 became hyperplastic, and part of them progressed to benign gonadal stromal tumors, consisting of vacuolated and lipid-rich α-inhibin–positive tumor cells without cytologic atypia, reaching a size of up to 3 cm in diameter (Fig. 3D-F).

Figure 3.

Male control groups. (A-C) Morphology of testicular transplants in male animals of CG 8 6 months after transplantation. As in female recipients, testicular grafts were vital and presented as white to yellow spots in the right half of the liver (A) (stereomicroscopic aspect). The surrounding liver parenchyma was not altered (B) (PAS reaction), and the grafts were exclusively composed of Sertoli and Leydig cells, while the seminiferous tubules were fibrotic and devoid of germ cells (C) (PAS reaction). Glycogen content and cytology of the adjacent hepatocytes were normal. (D-F) Development of large gonadal stromal tumors in the grafts of CG 8. (D) Unstained liver slice (24 months after transplantation); two normal-sized grafts are visible in the left part of the image. The right part is dominated by a large, yellowish-white gonadal stromal tumor. The adjacent liver parenchyma surrounding the normal-sized transplants and the tumor is not altered. (E) This large stromal tumor (30 months after transplantation) has occupied a major part of the liver and shows cystic changes and areas of hemorrhagia. Histologically (F), these tumors were composed of gonadal stromal cells arranged in nests and cords with large vacuolated cytoplasm, reflecting an increased synthesis in lipid hormones (PAS reaction). No nuclear polymorphism was noted, and mitotic activity was low. The adjacent hepatocytes still showed no morphological alterations. The width of each panel represents: (B) 2.3 mm; (C) 700 μm; (D) 9.5 mm; (F) and 360 μm; (A) and (E) show a macroscopic overview.

Table 2. Serum Levels of Sex Steroid and Gonadotropic Hormones
Experimental GroupFollicle-Stimulating Hormone (ng/mL)Luteinizing Hormone (ng/mL)Estradiol (pg/mL)Progesterone (ng/mL)Testosterone (ng/mL)
  • NOTE. Data are expressed as the mean value ± SD. The number of tested animals appears in parentheses.

  • Abbreviations: ND, not done.

  • *

    Significantly different to data of the same experimental group at 24 months. Significantly different to data of the same time group: a, CG 1; b, CG 2; c, CG 3; d, CG 4; e, CG 5; f, CG 6; g, CG 7; h, CG 8; i, CG 9; j, CG 10; k, CG 11.

 6 months18.87 ± 2.86 (9)* abcdefijk1.220 ± 0.209 (9) abdejk68.85 ± 6.19 (9)* cdik54.55 ± 6.64 (9) ck0.122 ± 0.096 (9) k
 24 months6.09 ± 1.30 (7) bj0.883 ± 0.85 (8) ej42.69 ± 7.28 (7) ac71.24 ± 14.69 (8)0.107 ± 0.100 (8)
CG 1     
 6 months5.58 ± 0.69 (10) bdfjh0.493 ± 0.089 (6) bfj87.51 ± 8.85 (10)* bdefijk88.31 ± 21.06 (10)ND
 24 months6.23 ± 0.95 (10) bj0.595 ± 0.079 (7) bcj63.15 ± 4.27 (10) bdfgik123.15 ± 39.50 (10)ND
CG 2     
 6 months54.13 ± 5.32 (10) cdeik3.461 ± 0.320 (10) cdefhik51.83 ± 6.51 (10) c47.92 ± 5.71 (10) ckND
 24 months31.82 ± 4.14 (8) cdeghik1.497 ± 0.290 (8) cdefghik37.01 ± 4.47 (9) c53.61 ± 9.07 (9)ND
CG 3     
 6 months5.22 ± 0.63 (10) dfhij1.006 ± 0.267 (9)*90.49 ± 6.95 (10)* defijk102.73 ± 17.05 (10) dejk0.030 ± 0.017 (10) ik
 24 months5.87 ± 1.41 (9) gij0.642 ± 0.055 (9) dij61.79 ± 4.26 (9) dfik103.08 ± 40.61 (9)0.018 ± 0.018 (9) ik
CG 4     
 6 months12.86 ± 0.79 (10)* efijk0.347 ± 0.046 (7) fj42.67 ± 2.10 (10) h53.78 ± 8.55 (10)ND
 24 months9.14 ± 1.13 (9) eghjk0.313 ± 0.057 (6) i42.99 ± 1.33 (9)39.75 ± 5.71 (9)ND
CG 5     
 6 months6.58 ± 0.46 (10) fhij0.370 ± 0.042 (3) fj51.44 ± 28.27 (10) h50.73 ± 6.41 (10) kND
 24 months6.38 ± 0.35 (10) gj0.285 ± 0.064 (5) hi46.12 ± 7.57 (9)41.64 ± 5.45 (10)ND
CG 6     
 6 months44.71 ± 1.70 (4) ik1.962 ± 0.344 (4)* fijk49.79 ± 2.12 (4)48.58 ± 8.93 (4) kND
 24 months20.96 ± 10.44 (10)0.653 ± 0.130 (9) j39.63 ± 6.49 (10)56.14 ± 10.84 (10)ND
CG 7     
 6 months2.94 (1)0.490 (1)72.41 (1)52.56 (1)ND
 24 months2.37 ± 0.31 (9) ij0.596 ± 0.167 (6)47.20 ± 6.39 (9)69.59 ± 16.77 (9)ND
CG 8     
 6 months51.98 ± 33.97 (4)*1.755 ± 0.971 (4)*68.37 ± 11.56 (4)60.57 ± 7.35 (4) jk0.064 ± 0.039 (4) k
 24 months4.45 ± 0.75 (16) j0.603 ± 0.071 (14) j79.02 ± 23.84 (16)63.44 ± 10.88 (16)0.470 ± 0.404 (9)
CG 9     
 6 months4.20 ± 0.60 (4) j0.655 ± 0.248 (3) j48.69 ± 1.94 (4)43.26 ± 10.11 (4)0.315 ± 0.185 (4)
 24 months7.13 ± 1.42 (7) j0.741 ± 0.087 (6)42.78 ± 5.03 (7)42.79 ± 6.58 (7)0.162 ± 0.042 (7) j
CG 10     
 6 months48.99 ± 2.34 (3) k3.810 ± 0.305 (3) k47.44 ± 2.82 (3)30.40 ± 7.11 (3)0.000 ± 0.000 (3) k
 24 months38.67 ± 11.04 (6) k2.936 ± 1.069 (6)67.04 ± 22.94 (5)47.91 ± 10.69 (6)0.000 ± 0.000 (6)
CG 11     
 6 months4.80 ± 0.77 (4)0.432 ± 0.115 (3)46.36 ± 0.88 (4)24.78 ± 12.39 (4)*0.501 ± 0.102 (4)*
 24 months5.04 ± 1.37 (8)0.464 ± 0.083 (4)48.24 ± 2.83 (8)66.61 ± 13.90 (8)0.176 ± 0.073 (8)

Foci of Altered Hepatocytes

In the MG, 6 months after transplantation, the majority of the liver acini located downstream of the transplanted ovarian fragments showed morphological alterations when compared with the adjacent parenchyma. Stereomicroscopically, these FAH were darker than the surrounding unaltered liver tissue (Fig. 1B) as a result of decreased glycogen content of the hepatocytes, as detected histochemically. In addition, at the histological level, altered hepatocytes showed cytoplasmic amphophilia, enlarged nuclei with decondensed chromatin and prominent nucleoli, and an increase in mitotic and apoptotic acitivity. Moreover, the typical cell plate pattern of liver parenchyma was replaced in these FAH by thickened hepatocellular trabeculae consisting of more than two cell layers (Fig. 1G). Ultrastructurally, the hepatocytes were mainly characterized by a loss of glycogen and an increase in mitochondria and peroxisome density. FAH did not develop in the left liver part of any animal in the MG nor in the livers of any male or female animals in the CGs, irrespective of whether ovarian, testicular, or no tissue was transplanted. It is particularly noteworthy that toremifene treatment in CG 4 completely prevented the development of FAH in these animals (Fig. 2B-D). Moreover, the liver parenchyma in the toremifene-treated animals of CG 4 and CG 5 was normal; in particular, no inflammation, fibrosis, cirrhosis, or any type of focal hepatocellular alteration was noted.

After 12, 18, 24, and 30 months, the majority of FAH persisted and showed a gradual progession of the architectural disorder. In addition, PAS reaction revealed a mild increase in glycogen content of single altered hepatocytes. Otherwise, the cytological aspect of the preneoplastic hepatocytes was largely constant and did not change significantly with increasing time if the lesion showed no neoplastic transformation. None of the CGs showed any of these alterations.

Development of Hepatocellular Neoplasms

The first HCA emerged 18 months posttransplantation (Table 1, Fig. 1C). Twenty-seven percent of the MG animals, but no CG animals, exhibited a HCA in the right part of the liver. HCAs showed a trabecular pattern and compressed the surrounding parenchyma by expansion, distorting the regular form of the liver acini. In general, the cytological aspect of the hepatocytes in the HCAs was very similar to that of amphophilic FAH, with relatively more variability in the glycogen content of the neoplastic cells when compared with the uniformly glycogen-poor hepatocytes of early FAH. Mitotic activity was moderately increased when compared with the FAH.

After 24 months, 42% of the MG animals displayed HCAs, and two control animals also showed sporadic HCAs. In addition, HCCs were observed for the first time (Fig. 1D-E) and were exclusively found in MG animals: 74% of them exhibited a HCC, and five showed an additional HCA. All HCCs were restricted to or clearly originated from the right part of the liver (Fig. 1D). Some of these tumors led to spontaneous death. The tumors measured up to 30 mm in diameter, showed a trabecular architecture with pseudoglandular growth, and grossly infiltrated the surrounding liver tissue. Several HCCs showed nodule-in-nodule changes indicating a transformation from HCAs. The cytological aspect of the preneoplastic hepatocytes was largely preserved but carcinoma cells showed more mitoses and cellular pleomorphism (Fig. 1H). Ovarian graft tissue was sometimes demonstrable within them (Fig. 1E). Ultrastructurally, a very unusual close spatial relationship between smooth endoplasmic reticulum and mitochondria was noted (Fig. 4).

Figure 4.

Electron micrograph of a HCC cell (MG, 24 months after transplantation). The number of mitochondria is high. Some peroxisomes are visible, and glycogen is depleted. Note the highly atypical localization of abundant smooth endoplasmic reticulum between mitochondria. The width of the image represents 7.9 μm.

Thirty months after transplantation, all 4 MG animals displayed a HCC originating from the right liver part and measuring 15 to 48 mm in diameter. The animal with the largest tumor developed multiple pulmonary metastases (Fig. 1I). After 24 and 30 months, the number of tumor- and carcinoma-bearing rats was significantly higher in the MG when compared with the respective CGs.

TGF-α Immunohistochemistry

Hepatocytes of unaltered liver tissue did not show any TGF-α expression. After 6 months, the FAH in the main group were also negative for TGF-α; however, after 12 months, we observed the first TGF-α–positive hepatocytes within the FAH (Fig. 5A). They were often located near the transplants. The proportion of TGF-α–positive hepatocytes in the FAH increased gradually, and most of the HCAs and HCCs showed moderate or strong TGF-α overexpression (Fig. 5B-C).

Figure 5.

TGF-α immunohistochemistry. (A) TGF-α–positive hepatocytes begin to develop in preneoplastic hepatocytes within the FAH not before 12 months after transplantation. The vast majority of hepatocellular neoplasms show TGF-α overexpression. (B) The small hepatocellular adenoma (center) adjacent to the transplant (upper edge) shows a uniformly strong overexpression (18 months after transplantation) and is thus sharply contrasted with the unstained adjacent liver parenchyma. (C) TGF-α staining in this large carcinoma 24 months after transplantation (left two thirds) varied slightly but was strongly positive in the majority of the cells, whereas the adjacent nonneoplastic hepatocytes (right edge) were TGF-α–negative. The width of each panel represents: (A) 1.1 mm; (B) 2.9 mm; and (C) 6.8 mm.

Levels of Sex Steroid and Gonadotropic Hormones

Aortic Blood.

The serum levels of follicle-stimulating hormone, luteinizing hormone, estradiol, progesterone, and testosterone are given in detail in Table 2. As expected, 6 months after transplantation, estradiol levels of the MG were significantly lower than those of the completely untreated CG 3 and the noncastrated CG 1, but higher than the solely ovariectomized rats of CG 2. Gonadotropin levels—particularly those of follicle-stimulating hormone—were inversely altered in the respective groups; thus, the levels were considerably elevated in the MG, indicating that the intrahepatically transplanted ovarian fragments synthesized estradiol to such an extent that the metabolizing capacity of the hepatocytes was surpassed, and that the systemic estradiol level increased but stayed clearly subnormal. After 24 months, the levels of follicle-stimulating hormone in the MG returned to normal, although estradiol levels were still systemically subnormal. The situation in the male animals of the testicular transplantation experiment was similar (Table 2).

Intrahepatic Estradiol Concentrations.

The intrahepatic estradiol level did not differ much between the left and right liver lobe in the untreated CG 3 animal (left and right part 103.3 pg/mL and 147.3 pg/mL, respectively) and in the ovariecomized CG 2 rat (left and right part, <5.3 pg/mL and 14.6 ± 4.7 pg/mL, respectively). However, they reflect the hypoestrogenemic situation after castration in the latter rat, because they were more than 10-fold lower than in the control animal.

In the MG animals, however, the high secretory activity of the stimulated grafts resulted in enormous differences of local intrahepatic E2 levels between the right and left liver lobe. Whereas the values in the transplant-free left liver lobe were between <5.3 pg/mL and 75.8 pg/mL, their range was 419.4 pg/mL to >2,285 pg/mL in the liver parenchyma neighboring the grafts. Thus, the difference between the right and left part in the 4 MG animals was at least 20.9-fold, 26.3-fold, 57.7-fold, and 92.4-fold, respectively.

Hepatic Vein.

Estradiol concentration in the venous blood that flows out of the liver was 36.7 ± 5.2 pg/mL in the ovariectomized CG 2 rats and 147.9 ± 20.2 pg/mL in the untreated control animals of CG 3. In the MG, the median estradiol concentrations in the hepatic vein blood were between 114.7 ± 33.2 pg/mL and 236.2 ± 36.5 pg/mL (with one single measurement showing a value of >2,583 pg/mL). Thus, the values of the mixed venous liver blood were between the individual intrahepatic levels of the right and left liver half.


This study shows that transplantation of ovarian fragments into the livers of previously ovariectomized rats leads to preneoplastic hepatocellular alterations that may develop into HCA and HCC 18 to 30 months after transplantation. The complex experimental setting consisting of 11 control groups and the consideration of the morphology of the transplants, the anatomical localization of the carcinogenic process, and the hormone levels—in particular the intrahepatic estradiol levels—allow for the following mechanistic conclusions.

Nonstimulated Ovarian Fragments Do Not Induce Tumors.

Stimulation of the intrahepatic grafts by gonadotropins is an important prerequisite for the carcinogenic process. Gonadotropic stimulation was reached by way of castration in the MG animals and is documented by the serum hormone levels (Table 2); the graft morphology, showing all stages of follicle maturation; and the morphology of sex steroid–dependent tissues (uteri, vaginal smears) (supplementary data). Stimulation of the grafts is particularly well demonstrated by the very high intrahepatic estradiol levels in the vicinity of the grafts in the MG animals, which sharply contrasted to the levels of the left liver part and the control animals. Only in the MG animals did the liver acini that drain the blood from the transplantated grafts show distinctive hepatocellular alterations that were observed previously17 and are now proven to be preneoplastic as they gave rise to hepatocellular neoplasms in this study. Animals that received transplants but were not castrated did not show sufficient gonadotropin secretion, owing to an intact feedback inhibition of the pituitary gland by the intraabdominal ovaries (CG 1). Hence, the intrahepatic grafts were not sufficiently stimulated and did not induce hepatocellular carcinogenesis.

Foci of Altered Hepatocytes and Hepatocellular Tumors Develop Only in the Liver Acini Downstream of the Transplants in the Main Group.

This observation suggests that the factors altering the liver parenchyma originate from the ovarian grafts. This conclusion is supported by the finding that the hepatocellular alterations are strictly confined to those liver acini located downstream of the transplanted ovarian fragments from which they drain blood. In neighboring acini, even the first plate of hepatocytes remained completely unaltered. This strict spatial relationship of hepatocellular alterations and ovarian transplants, and the finding that the hepatocellular changes develop only in the MG animals with an intact gonadotropic stimulation, strongly suggest that it is indeed the high local level of estradiol that is the most relevant factor for the development of the amphophilic FAH. The existence of a hyperestrogenemic microenvironment in the altered liver acini has been particularly well demonstrated by the observation of extremely high intrahepatic estradiol levels.

Stimulated Ovarian Fragments Do Not Induce Tumors When the Estrogen Receptor Is Blocked by Toremifene.

There is no evidence in the literature for a direct tumorigenic potential of endogenous estradiol on the liver in vivo. In general, estrogens are known to induce tumors via both genotoxic and nongenotoxic mechanisms.8–11 Several facts strongly indicate that the primary trigger leading to the formation of preneoplastic altered liver acini in this model is nongenotoxic and is mediated by the estrogen receptor after direct binding of estradiol or previous activation of more reactive metabolites:

  • 1The formation of the lesions already begins 1 to 3 weeks after transplantation in all animals and, moreover, leads to virtually uniform alterations within the FAH. Genotoxic effects would lead to a greater diversity in alterations of the preneoplastic cells.
  • 2Carcinogenesis in this model shows several similarities to that induced by peroxisome proliferators18–20 acting through the peroxisome proliferator–activated receptor α, which belongs to the superfamily of steroid hormone receptors.28, 29
  • 3Carcinogenesis is completely prevented from the very beginning (i.e., the formation of the preneoplastic FAH), when the animals were additionally treated with antiestrogenic toremifene. Like tamoxifen, toremifene acts predominantly antiestrogenically, mainly by blocking the estrogen receptor (normally expressed in the FAH; data not shown), but toremifene, in contrast to tamoxifen, has no tumorigenic potential on the liver.30–34

Toremifene-induced tumor prevention not only highlights the contribution of the estrogen receptor, but clarifies that other secretory products of the ovarian grafts, such as progesterone or α-inhibin, although possibly modulating estrogen effects, are not the primary triggers of carcinogenesis. The molecular events that are important for carcinogenesis in the complex intracellular network related to estradiol action have yet to be investigated.

TGF-α Is Upregulated in Altered Foci Downstream of Transplants in the Main Group and Is Strongly Positive in Tumors.

It is clear that, in the course of the carcinogenic process, other factors that facilitate the progression and possibly also the transformation of late-stage preneoplasias into hepatocellular neoplasms become increasingly important in this model. This is corroborated by the serum hormone levels at 24 months that indicate a diminishing influence of the initially decisive hormonal stimulation within the liver parenchyma and also by a similar development in late-stage lesions and tumors in another endocrine hepatocarcinogenesis model that is based on the transplantation of pancreatic islets into diabetic animals and in which a gradual independence after several months of the primarily decisive hyperinsulinism was demonstrated.35, 36 Similar to that model, we found TGF-α overexpression in late-stage preneoplasias and tumors, suggesting that TGF-α may also be of importance for the progression or neoplastic transformation of late preneoplasias in this model, particularly considering the well-known comitogenic effects of TGF-α with estradiol.13

The Morphology of Preneoplastic Foci and Hepatocellular Tumors Differs Substantially From Other Hormonal Models.

The results of this study fit into our concept that long-term high concentrations of certain hormones (e.g., thyroxin, insulin, and estradiol) possess hepatocarcinogenic potential.35–43 These models share a common trait in that the formation of preneoplastic lesions is at first a result of metabolic adaptation to high local hormone concentrations before additional nongenotoxic and genotoxic effects may lead to neoplastic transformation. Low-number transplantation of pancreatic islets into the livers of diabetic rats induces another phenotype of FAH—the clear-cell (glycogen-storing) focus that progresses stepwise into HCA and HCC35, 36, 38–43—while thyroxin induces other alterations37 that are nearly identical to the so-called “amphophilic” FAH mentioned above. Although the type of FAH that is induced in this study shares many similarities with the amphophilic focus, there are also some deviations, which are described in detail in our previous short-term study.17 Thus, the unique pattern of hepatocellular alterations in this model may represent typical estrogen effects; however, this remains to be proven, because it has never been investigated outside of these two studies. Nevertheless, it can at least be said that each type of endocrine stimulation leads to another distinctive type of FAH, illustrating a different influence of the respective hormones on the hepatocellular metabolism. In this regard, it is also interesting that similar effects on hepatocytes are noted in different models of chemical hepatocarcinogenesis. Thus, one can in turn interpret some of these effects of chemical carcinogens as an imitation of impaired hormonal action.44

Stimulated Testicular Fragments Secreting Testosterone Do Not Induce Tumors in Male or Female Rats.

The absence of any morphological changes after testicular graft transplantation was unexpected. Testicular transplants lost their germ cells within the first 3 months after transplantation via apoptotic elimination and thus were exclusively composed of Sertoli and Leydig cells, possibly indicating an impaired metabolic situation or insufficient stimulation. However, these animals, in contrast to solely castrated rats, did show detectable testosterone levels in aortic blood; therefore, the hepatocytes downstream of the grafts must have been exposed to considerable testosterone levels, which is supported by the observation of massively testosterone-producing, graft-derived stromal tumors in some animals. The lack of influence of testosterone on hepatocarcinogenesis in this study is unclear but at least demonstrates that factors that may be related to the transplantation procedure per se (e.g., ischemia) are not relevant for carcinogenesis in this context.

In conclusion, we established a model of endocrine hepatocarcinogenesis based on the effects of high local concentrations of endogenous estrogens after intrahepatic transplantation of ovarian tissue. The carcinogenic process is triggered by nongenotoxic, receptor-mediated estrogen mechanisms; thus, this model serves as an example of hepatocarcinogenesis, beginning with a hormonally induced dysregulation of the cell metabolism of FAH that hold a preneoplastic potential and that may gradually transform to HCA and HCC.


The authors wish to thank Gabriele Becker, Jörg Bedorf, Danuta Chrobok, Mariana Dombrowski, Mathilde Hau-Liersch, and Kirsten Herrmanns for technical assistance; Yvonne Fischer and Kurt Rüdel for animal care; Thomas Jonczyk-Weber for photographic work; and Bernd Wüsthoff for editing the manuscript.