• Open Access

Insulin-independent promotion of chemically induced hepatocellular tumor development in genetically diabetic mice


To whom correspondence should be addressed. E-mail: ghlee@kochi-u.ac.jp


Diabetes mellitus has been proposed as an epidemiological risk factor for human liver cancer development. One reasonable possibility is that this is attributable to hyperinsulinemia compensatory for obesity-related insulin resistance. However, diabetes mellitus is a complex disease with multiple abnormal conditions essentially caused by hyperglycemia. Therefore, it is not evident whether hyperinsulinemia is prerequisite for the elevated cancer risk. To gain a clue to answer this question, we characterized chemically induced hepatocarcinogenesis in diabetic model mice genetically deficient for insulin. Akita inbred mice originating from the C57BL/6 strain carry a heterozygous germline mutation of the insulin II gene and suffer from inherited insulin deficiency and diabetes in an autosomal dominant manner. They were mated with normal C3H/HeJ mice with high sensitivity to liver carcinogenesis and the resultant F1 littermates, which were either normal or insulin deficient, were exposed to diethylnitrosamine and induced hepatocellular tumors were evaluated for number, size, proliferative activity, and apoptosis. Unexpectedly, both mean and total volumes of hepatocellular tumors in the insulin-deficient animals were more than twofold larger than those in the normal controls, with no significant difference in tumor number. The tumors in insulin-deficient mice showed a significantly lower frequency of apoptosis but no alteration in cell proliferation. In conclusion, our results indicate that insulin-independent liver tumor promotion occurred in diabetic mice. Clearly, insulin-independent mechanisms for the human case also deserve consideration. (Cancer Sci 2009)

Although persistent hepatitis B or C viral infection is the rate-limiting risk factor for human liver carcinogenesis in most cases,(1,2) the process can be significantly modified by other agents, one of which is diabetes mellitus (DM). Epidemiological studies have consistently confirmed DM, mostly of type 2, to be associated with approximately twofold elevation in the frequency of hepatocellular carcinonomas.(3–5) This is a serious social issue, as the number of DM patients is now rapidly increasing, particularly in developed countries. Similar epidemiological data have also been reported for cancers of the pancreas, endometrium, colorectum, and breast,(6) while data for the lung, for instance, are controversial,(7,8) suggesting organ-specific influences. As one recent study proposed that DM promotes human hepatocarcinogenesis associated with chronic hepatitis C, but not hepatitis B, DM-related liver cancer risk may also depend on the type of viral infection.(9)

The mechanism of liver cancer promotion by DM is not well understood, mainly because DM of type 2 is a consequence of a complex pathogenesis involving both obesity-related insulin resistance and eventual impaired insulin secretion. Nevertheless, many investigators have proposed that compensatory hyperinsulinemia associated with insulin resistance seen in pre-DM or early DM populations with obesity should play a role in promoting cancer development.(10) This is a reasonable idea because insulin has been linked with promotion of cellular growth, the same as insulin-like growth factors.(11) In fact, one epidemiological report from the USA described overall cancer risk to be more pronounced in insulin-resistant, pre-DM people with hyperinsulinemia relative to non-compensatory, overt DM patients.(12) On the other hand, it is also worth paying attention to one recent report of a Korean DM cohort indicating that cancer risk is simply influenced by fasting blood glucose in a dose-dependent manner, uncorrelated with the body mass index, a measure of insulin resistance.(13) This casts doubt on the supposed direct action of insulin on promotion of cancers. Although the controversy could be approached by evaluating the cancer risk of type 1 DM patients with insulin deficiency, this is hampered by the shortage of cases and bias from prolonged insulin therapy.

In terms of liver carcinogenesis, a series of studies by Dombrowski et al. demonstrated that DM model rats with insulin deficiency develop spontaneous liver tumors after intrahepatic transplantation of pancreatic islets.(14,15) Liver tumor development was, however, not observed when a high numbers of islets was transplanted. These experimental results suggest that under DM conditions, a locally hyperinsulinemic milieu, which may be more effectively achieved by low rather than high number islet transplantation, is hepatocarcinogenic. Although this is a remarkable contribution to understanding the relationship between insulin and liver cancer, the hepatic islet transplantation does not necessarily reflect the natural insulin environment in human DM patients. While rats and hamsters treated with streptozotocin, a toxicant against pancreatic β cells, develop severe DM followed by occasional liver tumors,(16,17) streptozotocin itself is a mutagen,(18) so it is not clear whether the liver tumor formation is an outcome of streptozotocin-induced genetic damage or DM.

The Akita mouse suffers from autosomal dominantly inherited DM simply caused by a dominant-negative effect of a mutated Insulin II (Ins2) gene leading to insulin deficiency.(19,20) Although the pathogenesis is similar to that of human type 1 DM, the animals survive without insulin therapy. Thus, it is an ideal model to investigate the biological effects of insulin deprivation in vivo. In the present study, to gain clues to understanding the relationships among DM, insulin, and cancer, we characterized the effects of the Akita mutation on chemically initiated hepatocarcinogenesis. As Akita mice originated from the C57BL/6 strain,(19) which is very resistant to liver tumors, we used F1 mice generated by crossing Akita mice with the liver tumor-susceptible strain C3H/HeJ (C3H).(21,22) If liver cancer risk largely depends on hyperinsulinemia rather than DM or hyperglycemia as a consequence, the Akita mutation would be expected to have no effect or even a negative effect on hepatocarcinogenesis. However, we obtained the opposite result.

Materials and Methods

Liver tumor induction in mice.  All experiments were approved by the Kochi University Animal Experiment Committee and carried out in accordance with the Guidelines for Animal Experiments at Kochi University. The Akita mouse is a mutant substrain of the C57BL/6 inbred mouse. It possesses a germline mutation in one of the two Ins2 gene alleles, which causes autosomal dominant inheritance of insulin deficiency.(19,20) For convenience, the normal and mutated Ins2 alleles are hereafter designated Ins2 and Akita, respectively. Male C57BL/6Ins2/Akita mice (i.e. Akita mice), and female C3HIns2/Ins2 mice (i.e. normal C3H mice) were mated and male littermates delivered were used for the experiment because male mice are generally more sensitive than female mice to both DM and liver carcinogenesis.(19,23) Approximately half of the littermates should be insulin-deficient (C3H × C57BL/6)F1Ins2/Akita mice and the rest should be normal (C3H × C57BL/6)F1Ins2/Ins2 mice. Because all of the F1 mice were genetically identical except for the Ins2 genotype, (C3H × C57BL/6)F1Ins2/Ins2 mice served as normal controls. At 14 days of age, 39 male F1 mice were intraperitoneally injected with 5 μg/g bodyweight of diethylnitrosamine (WAKO, Osaka, Japan) dissolved in physiological saline, as previously described.(24) At this time, the Ins2 genotype of each F1 mouse was not known. However, the infant (C3H × C57BL/6)F1Ins2/Akita and (C3H × C57BL/6)F1Ins2/Ins2 mice were supposed to be equivalent in mean bodyweight, as abnormal weight loss of Akita mice becomes evident only after 18 weeks of age.(19) All of the dietylnitrosamine-injected mice were weaned at 21 days of age and allowed free access to basal diet (CEII; CLEA Japan, Tokyo, Japan) and deionized water. At 32 weeks of age, the animals were tested for sugar in urine with test tapes (Pretest; WAKO) to discriminate insulin-deficient (C3H × C57BL/6)F1Ins2/Akita mice and normal (C3H × C57BL/6)F1Ins2/Ins2 mice as controls, and then killed by cervical decapitation. For each mouse, the body, epididymal white fat, and liver were weighed. Representative slices were cut from all liver lobes and fixed in buffered 10% formalin. Portions of liver tumors larger than 5 mm in diameter and corresponding normal liver tissue were frozen on dry ice. Before death, blood samples of some randomly selected animals were drawn from femoral veins under ether anesthesia, and used for measurement of blood glucose levels with a portable blood-sugar testing device (Medisafe; Terumo, Tokyo, Japan) and preparation of serum. The frozen tissues and sera were stored at −70°C until use.

Stereological quantification of hepatocellular proliferative lesions.  The formalin-fixed liver slices of each mouse were embedded in paraffin and sections were made. Total number, mean, and total volumes, and size distribution of focal hepatocellular proliferative lesions were estimated according to the stereological method of Enzmann et al. with modification of size classes.(25) The lesions were identified on hematoxylin–eosin-stained specimens and lesion profile diameters were recorded using routine light microscopy and an eyepiece micrometer. To ensure reproducible identification of lesions, those smaller than 160 μm in profile diameter were not included in the analysis. The mean area examined per liver was 2.3 ± 0.1 cm2 (mean ± SEM).

The method by Enzmann et al. makes some assumptions on the size and distribution of the lesions. Such information can be obtained from their original report.(25)

Scoring of fatty change within liver tumors.  Scoring of fatty change was done with hematoxylin–eosin-stained sections measuring the percentage of area occupied by cytoplasmic vesicular fat droplets within each liver tumor profile, as follows: score 0, <5%; score 1, 5–33%; score 2, 34–66%; and score 3, >66%.

Immunohistochemistry.  With formalin-fixed paraffin sections, immunohistochemical detection of the Ki-67 protein, 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 (Pfkfb3) protein, and single-stranded DNA (ssDNA) was carried out using the avidin–biotin complex method as previously described,(26) using PBS-diluted primary antibodies as follows: rat monoclonal antimouse Ki-67 antibodies (Dako, Glostrup, Denmark), rabbit polyclonal antimouse Pfkfb3 antibodies (ABGENT, San Diego, CA, USA); and rabbit polyclonal anti-ssDNA antibodies (Dako). For Ki-67 and Pfkfb3, the deparaffinized sections were heated at 100°C for 30 min in acidic citrate buffer (pH 6) before application of primary antibodies.

Ki-67 labeling index.  Expression of the Ki-67 protein is a reliable indicator of the cell proliferative activity of mouse tumor cells.(27) In the present study, Ki-67 labeling indices were expressed as the percentage of tumor cells with immunohistochemically Ki-67-positive nuclei under a light microscope. At least 500 tumor cells were examined for each tumor.

Apoptotic index.  ssDNA staining of paraffin sections is a reliable immunohistochemical marker for specific detection of apoptosis.(28) All ssDNA-positive apoptotic figures observed in the tumor area were counted and the apoptotic index was expressed as the number of apoptotic figures per 1 cm2 of pooled tumor profile area, measured using the VS-100 virtual slide system (Olympus Corporation, Tokyo, Japan). When multiple apoptotic bodies were tightly assembled, they were considered to compose a single apoptotic figure.

Western blot analysis.  Western blot analysis of the Pfkfb3 or β-actin proteins was carried out using the antimouse Pfkfb3 antibodies used for the immunohistochemistry or mouse monoclonal anti-β-actin antibodies (Sigma, St Louis, MO, USA) essentially as previously described.(26) The proteins were detected using the enhanced chemiluminescence system (GE Healthcare, Buckinghamshire, UK) and X-ray film. Densitometrical quantification of signal intensity was carried out with the ImageJ 1.33u software freely distributed by the National Institutes of Health (Bethesda, MD, USA). Relative Pfkfb3 levels adjusted for β-actin content were calculated according to the densitometric data.

ELISA.  Serum concentrations of specific mouse proteins were measured using commercially available ELISA kits for mouse insulin (Morinaga Institute of Biological Science, Kanagawa, Japan), insulin-like growth factor (Igf) 1 (R & D Systems, Minneapolis, MN, USA), insulin-like growth factor binding protein (Igfbp) 3 (BioVendor, Modrice, Czech), adiponectin (Otsuka Pharmaceutical Company, Tokyo, Japan), and leptin (Morinaga Institute of Biological Science). Quantification was done according to the provided manufacturer’s manuals.

Experimental insulin therapy.  Human insulin (Novolin N; Novo Nordisk Pharma, Tokyo, Japan) was subcutaneously injected into the scapular regions of five insulin-deficient mice at 0800 and 2000 hours every day. The dosage of each injection was 4 U per mouse. For five control mice, physiological saline was injected instead of insulin. After 5 days, all mice were killed by cervical decapitation. Their livers were resected, fixed in buffered formalin, and analyzed.

Statistical analysis.  Because a normal distribution of the analyzed variables was not guaranteed, the two-sided exact Mann–Whitney U-test, a non-parametric counterpart for Student’s t-test, was used for statistical comparison of mean values. For confirmation, the parametric Student’s t-test and Welch’s t-test were carried out simultaneously. Differences were considered statistically significant when P-values less than 0.05 were obtained for all of the three tests. In terms of judgment on significance, there was actually no inconsistency between the results of the three statistical tests. Only the P-values for the Mann–Whitney U-test are presented.


Modulation of liver tumor development by insulin deficiency.  During the experiment, only one insulin-deficient mouse died, for an unknown reason. When killed, the mean blood glucose levels, as expected, were 3.8-fold higher in insulin-deficient (C3H × C57BL/6)F1Ins2/Akita mice than in the control (C3H × C57BL/6)F1Ins2/Ins mice (Table 1). While the insulin-deficient mice demonstrated significantly lower mean bodyweights and epididymal fat weights, their mean liver weights were significantly higher than in control mice by approximately 30%, apparently due to a heavier burden of liver tumors (Table 1).

Table 1.   Mean bodyweights, epididymal fat weights, liver weights, and blood glucose levels of control (C3H × C57BL/6)F1Ins2/Ins2 and insulin-deficient (C3H × C57BL/6)F1Ins2/Akita mice 8 months after a single injection of diethylnitrosamine
 Control miceInsulin-deficient mice
  1. †Values are mean ± SEM. ‡Numbers in parentheses are numbers of mice analyzed. *< 0.001 compared with control mice.

Bodyweight (g)38.8 ± 0.8† (19)‡32.3 ± 1.1* (14)
Epididymal fat weight (g)0.45 ± 0.10 (19)0.06 ± 0.02* (14)
Liver weight (g)2.1 ± 0.1 (19)2.7 ± 0.1* (14)
Blood glucose level (mg/dL)191 ± 16 (8)730 ± 20* (8)

All of the diethylnitrosamine-injected mice developed multiple liver tumors irrespective of the Ins2 genotype. The gross appearance of the tumors was dependent on the insulin status. As shown in Figure 1, the liver tumors observed in control mice were generally whitish, so that they are easily recognizable against normal liver. In contrast, most of the tumors developing in insulin-deficient mice were dark brown, similar to the background normal liver, making their gross detection difficult. For this reason, we decided to evaluate liver tumor development with histological sections using an established stereological method.(25)

Figure 1.

 Gross appearance of (a) a control and (b) an insulin-deficient mouse liver with multiple tumors (scale bar = 1 cm). While most tumors in the control mouse liver are whitish and easily recognizable, many of those in the insulin-deficient mouse liver are dark in color and difficult to identify.

Histopathologically, hepatocellular proliferative lesions in the mouse liver may be categorized into foci of cellular alteration, adenomas, or carcinomas.(29) We combined these lesions as hepatocellular proliferative lesions in stereological analyses, as distinguishing between foci of cellular alteration and small adenomas is practically difficult and hepatocellular carcinomas comprised less than 1% of all the lesions. The results are summarized in Table 2. Although the mean lesion number per liver for insulin-deficient mice was slightly larger than that of control mice, the difference was not statistically significant. On the other hand, both mean and total volumes of the lesions were 2.2 times higher for insulin-deficient mice than for control mice. The estimated lesion size distribution also confirmed that insulin-deficient mice developed larger lesions than the controls (Fig. 2). The difference was most pronounced for lesions larger than 1.277 mm in diameter, which is equivalent to the size of a sphere 1090 × 106 μm3 in volume.

Table 2.   Development of diethylnitrosamine-induced hepatocellular proliferative lesions in control (C3H × C57BL/6)F1Ins2/Ins2 and insulin-deficient (C3H × C57BL/6)F1Ins2/Akita mice
 Control miceInsulin-deficient miceP-value
  1. †Values are mean ± SEM. P-value obtained by statistical comparison of the data for control and insulin-deficient mice.

Number of mice analyzed  19 19 
Mean number of lesions per liver325 ± 41†391 ± 540.5
Mean total volume of lesions per liver (cm3)0.38 ± 0.080.85 ± 0.100.0008
Mean volume of lesions (106 μm3)1160 ± 1802560 ± 220<0.0001
Mean volume of normal livers (cm3)1.67 ± 0.071.86 ± 0.060.06
Figure 2.

Size distributions of hepatocellular proliferative lesions developing in control or insulin-deficient mice. Lesion size is expressed as diameter. The size classes are according to the method of Enzmann et al.(25)

Although the mean normal liver volume was 10% larger for insulin-deficient mice, the difference did not quite reach significance (Table 2). In both control and insulin-deficient mice, the non-tumorous liver tissue was histologically unremarkable and devoid of cytotoxic damage, inflammatory reaction, and steatosis.

Proliferation and apoptosis of tumor cells.  Ki-67 and ssDNA labeling indices, which have been widely used as reliable measures for tumor cell proliferation and apoptosis respectively,(27,28) were immunohistochemically quantified. The results are summarized in Table 3.

Table 3.   Ki-67 labeling and apoptotic indices of liver tumors in control (C3H × C57BL/6)F1Ins2/Ins2 and insulin-deficient (C3H × C57BL/6)F1Ins2/Akita mice
 Control miceInsulin-deficient miceP-value*
  1. †Three tumors largest in profile area were selected from each mouse and analyzed.

  2. ‡Values are mean ± SEM. §The number of Ki-67-positive tumor cells per 100 tumor cells.¶The number of ssDNA-positive apoptotic figures per 1 cm2 of pooled tumor profile area. P-value obtained by statistical comparison of the data for control and insulin-deficient mice.

Number of mice analyzed  5    5 
Ki-67 analysis
 Number of tumors selected for analysis† 15   15 
 Mean profile area of selected tumors (mm2)1.6 ± 0.3‡4.3 ± 0.4<0.0001
 Mean Ki-67 labeling index of tumors (%)§3.1 ± 0.52.9 ± 0.30.8
Apoptosis analysis
 Apoptotic index (no. ssDNA-positive figures per cm2158.5 ± 35.228.5 ± 4.50.008

For the Ki-67 analysis, only the three tumors largest in profile area were selected from each of five insulin-deficient mice and five control mice, because for smaller tumors accurate evaluation of the labeling index was hampered by a shortage of countable tumor cells. The mean profile area of analyzed tumors was significantly larger for insulin-deficient mice than for control mice, implying that the selected tumors of insulin-deficient mice enlarged faster in general. Nevertheless, there was no significant intergroup difference in the mean Ki-67 labeling indices (Table 3).

The apoptotic frequency was expressed as number of ssDNA-positive apoptotic figures per 1 cm2 of pooled tumor profile area, as our previous study indicated that apoptotic bodies in diethylnitrosamine-induced mouse liver tumors are rare.(24) The results indicated that apoptotic frequency is five times higher for control mice relative to insulin-deficient mice (Table 3). Before this evaluation, we confirmed that there is no significant difference in mean cellular density between the tumors of insulin-deficient mice and those of control mice (data not shown).

Fatty change within liver tumors.  We realized that tumors in control mice more frequently show central fatty change characterized by cytoplasmic accumulation of microvesicular and macrovesicular fat droplets compared with those in insulin-deficient mice (Fig. 3a,b). This should reflect generally whitish gross appearance of the control mouse tumors. Thus, fatty change in each tumor was evaluated according to the scoring criteria described in the Materials and Methods section. Because the change tended to be more prominent in larger tumors, the scoring was done with reference to tumor profile diameter on the histological section.

Figure 3.

(a) Histology of a hepatocellular tumor without fatty change (fatty change score = 0) developing in an insulin-deficient mouse and (b) one with marked fatty change (fatty change score = 3) in a control mouse. An arrowhead indicates each tumor. Scale bar = 0.2 mm. (c) Two single-stranded DNA-positive apoptotic figures (arrows) observed in a tumor with marked fatty change (fatty change score = 3) in a control mouse.

As shown in Table 4, in both control and insulin-deficient mice, tumors sized 1 mm and over in profile diameter had significantly higher mean scores for fatty change than those smaller than 1 mm. Furthermore, in both size classes, tumors in control mice had a significantly higher mean score than those in insulin-deficient mice, indicating that insulin deficiency decreases fatty change in liver tumors.

Table 4.   Scoring of fatty change in liver tumors of control (C3H × C57BL/6)F1Ins2/Ins2 and insulin-deficient (C3H × C57BL/6)F1Ins2/Akita mice
Tumor profile diameterMean score† (mean ± SEM)P-value
Control miceInsulin-deficient mice
  1. †The score was defined according to percentage of tumor cells with vesicular fatty change as follows: score 0, <5%; score 1, 5–33%; score 2, 34–66%; score 3, >66%. ‡Numbers in parentheses are numbers of tumors analyzed. P-value obtained by statistical comparison of the data for control and insulin-deficient mice.

<1 mm0.3 ± 0.1 (39)‡ 0 ± 0 (49)0.0006
≥1 mm1.6 ± 0.2 (11)0.2 ± 0.1 (31)<0.0001

Although we observed ssDNA-positive apoptotic figures in larger tumors with fatty change (Fig. 3c), the general infrequency of apoptotic figures disabled us from analyzing an association between apoptosis and fatty change.

Expression of the Pfkfb3 protein.  One advantageous environment for tumors in insulin-deficient mice is ample blood glucose for glycolysis as a source of biological energy. Therefore, expression of Pfkfb3, a key enzyme upregulating glycolysis,(30,31) was examined by western blotting. Only tumors larger than 5 mm were analyzed because it was difficult to obtain sufficient fresh materials for the smaller tumors. As shown in Figure 4(a), the relative expression level of the Pfkfb3 protein adjusted by β-actin content was individually variable in both normal and tumor tissues, although the mean level for the tumors in insulin-deficient mice was 1.7 times higher than that for the tumors in control mice (Fig. 4b). This may reflect the known physiological fluctuation of Pfkfb3 gene expression in the liver.

Figure 4.

(a) Western blot analysis of 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 (Pfkfb3) protein in hepatocellular tumors and normal livers (N = normal liver; T = tumor). The relative Pfkfb3 level was determined with reference to the β-actin content of each sample. (b) Graphical depiction of the mean Pfkfb3 levels. The mean level for tumors from insulin-deficient mice was statistically different from that for tumors from control mice (= 0.03). Error bar = SEM.

The Pfkfb3 protein was also examined by immunohistochemistry using paraffin sections, as only tumors over 5 mm in diameter were used for the western blot study. Also, in insulin-deficient mice, macroscopic distinction of normal liver tissue from small liver tumors was practically difficult as mentioned earlier, possibly leading to sampling bias. As shown in Figure 5(a), practically all liver tumors lacking significant fatty change were stained stronger than their corresponding normal livers. In contrast, tumors with fatty change, which are typically over 1 mm in profile diameter, exhibited central staining defects (Fig. 5b–d). Although these features were independent of insulin status, more liver tumors in control mice exhibited the centrally defective staining pattern than in insulin-deficient mice, correlating with the mean fatty change score. The western blot study may have underestimated Pfkfb3 levels in the tumors, as only very large tumors with a propensity for fatty change were analyzed.

Figure 5.

Immunohistochemical detection of 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 (Pfkfb3) proteins in mouse livers with hepatocellular tumors. Arrowheads indicate tumors. (a,b) Insulin-deficient mice. (a) All tumors here are devoid of fatty change and uniformly stained for Pfkfb3. Note that the staining intensity for tumors is stronger than that for the normal liver. (b) A large tumor with a fatty change score of 3 indicated by a red arrowhead shows a centrally defective staining pattern. (c–e) Control mice. (c) Two relatively small tumors with a fatty change score of 1 exhibit a centrally defective staining pattern. (d) A very large tumor indicated by a red arrowhead with a fatty change score of 3 shows an extensive inner staining defect. (e) A higher magnification of the large tumor shown in (d) (red arrowhead). Note that the staining defect is not necessarily associated with vesicular fatty change at the cellular level. There is an aggregate of positive-staining cells in the central area (arrow). (a–d) Scale bar = 1 mm. (e) Scale bar = 0.2 mm.

Although the staining defect was centrally located associated with fatty change, both fat accumulating and non-accumulating tumor cells in the center were low in Pfkfb3 at the cellular level (Fig. 5e). Also, scattered small aggregates of Pfkfb3-positive tumor cells were common in the deficient areas (Fig. 5e).

Quantification of cancer-related serum proteins.  Increased serum Igf1, a structurally insulin-related hormone, has been implicated in human cancer risk as well as hyperinsulinemia.(32,33) Igfbp3, a major binding protein for Igf1, is considered a modifier of Igf1 bioavailability.(32,33) As indicators of obesity-associated hormone status, decreased adiponectin and increased leptin have been linked with the risk of some cancers, even if the results of studies on leptin are controversial.(34,35) Thus, we quantified serum concentrations of insulin, Igf1, Igfbp3, adiponectin, and leptin in the control and insulin-deficient mice by ELISA. The results are shown in Table 5. As expected, insulin was undetectable for the insulin-deficient mice, although it was well within the measurable range for the control mice. No significant differences in mean serum Igf1 and Igfbp3 levels were demonstrable between control and insulin-deficient mice. In clear contrast, the mean serum level of adiponectin for insulin-deficient mice was only one-third of that for control mice. For the insulin-deficient mice, leptin was essentially below the measurable level, while it was readily detectable in the control mice.

Table 5.   Serum concentrations of insulin, insulin-like growth factor (Igf) 1, insulin-like growth factor binding protein (Igfbp) 3, adiponectin, and leptin of control (C3H × C57BL/6)F1Ins2/Ins2 and insulin-deficient (C3H × C57BL/6)F1Ins2/Akita mice
ProteinControl miceInsulin-deficient miceP-value
  1. †Values are mean ± SEM. ‡Numbers in parentheses are numbers of mice analyzed. P-value obtained by statistical comparison of the data for control and insulin-deficient mice.

Insulin (ng/mL)2.7 ± 1.2† (8)‡0.0 ± 0.0 (8)0.0002
Igf1 (ng/mL)337 ± 35 (14)340 ± 25 (13)0.79
Igfbp3 (ng/mL)150 ± 13 (14)128 ± 13 (14)0.23
Adiponectin (μg/mL)11.5 ± 1.1 (14)3.6 ± 0.4 (14)<0.0001
Leptin (μg/mL)6.7 ± 0.9 (14)0.0 ± 0.0 (14)<0.0001

Effects of experimental insulin therapy on proliferation and apoptosis of liver tumor cells in insulin-deficient mice.  A pilot study of insulin therapy on the insulin-deficient mice indicated that they are extremely resistant to exogenous insulin when blood glucose was used as the parameter. There was also significant individual variation in insulin sensitivity. However, two subcutaneous injections of 4 U of human insulin per day were found to control blood glucose concentrations at approximately 50–300 mg/dL without hypoglycemic death. This injection amount is similar to that used for diabetic mice in a previous study.(36)

After 5 days of either the insulin therapy or saline injection to liver tumor-bearing insulin-deficient mice, the Ki-67 and ssDNA labeling indexes were evaluated for liver tumors (Table 6). No significant difference in Ki-67 labeling indices was observed between the insulin therapy and saline groups. Consequently, there was no evidence that exogenous insulin promotes cell proliferation in insulin-deficient mouse liver tumors.

Table 6.   Effects of experimental insulin therapy on Ki-67 labeling and apoptotic indices of liver tumors in insulin-deficient (C3H × C57BL/6)F1Ins2/Akita mice
  1. †Three tumors largest in profile area were selected from each mouse and analyzed. ‡Values are mean ± SEM. §The number of Ki-67-positive tumor cells per 100 tumor cells. ¶The number of ssDNA-positive apoptotic figures per 1 cm2 of total tumor profile area. P-value obtained by statistical comparison of the data for control and insulin-deficient mice.

Number of mice analyzed 5  5 
Ki-67 analysis
 Number of tumors selected for analysis†15 15 
 Mean profile area of selected tumors (mm2) 4.0 ± 0.6‡  4.0 ± 0.50.98
 Mean Ki-67 labeling index of tumors (%)§ 2.8 ± 0.4  2.9 ± 0.40.6
Apoptosis analysis
 Apoptotic index (number of ssDNA-positive figures per cm230.1 ± 4.4172.8 ± 45.00.008

On the other hand, tumors in the insulin-injected mice showed significantly higher frequency of apoptosis than the saline-injected mice by 5.7-fold (Table 6).


The present experimental study demonstrated that insulin-deficient mice develop significantly larger hepatocellular tumors than control mice, albeit without any significant difference in the tumor number. Thus, hereditary insulin deficiency may be considered a typical liver tumor promoter in terms of the two-stage concept of carcinogenesis, apparently contradicting the prevailing hypothesis that compensatory hyperinsulinemia observed in insulin-resistant type 2 DM patients is a primary cause of elevated cancer risk.(10)

Although we could not demonstrate a difference in Ki-67 index between control and insulin-deficient mouse liver tumors, the frequency of apoptosis with ssDNA-positive bodies was significantly lower in the latter. Considering the mean tumor cell density, the mean apoptotic indices for control and insulin-deficient mouse tumors were respectively equivalent to 6.6 and 1.2 apoptotic figures per 10 000 tumor cells. These values are comparable to previously reported apoptotic indices for mouse liver tumors and a similar magnitude of decrease in frequency may confer a significant growth advantage on mouse liver tumors.(24,37)

Our histological analysis revealed liver tumors to show a tendency for central fatty change with enlargement in size. Such change was here co-localized with low levels of Pfkfb3 protein, which promotes glycolysis as an energy provider and is highly expressed in many types of human cancers.(31) Notably, fatty change with lowered Pfkfb3 was relatively rare in the tumors of insulin-deficient mice. This might at least partly explain the infrequent apoptosis, as Pfkfb3 expression prevents such single-cell death in cultured human cancer cells.(38)

Physiological stimulation of glycolysis depends on insulin and glucose loads. However, under some conditions, insulin-independent glycolytic mechanisms may operate.(39) For instance, in mice deficient for insulin due to streptozotocin treatment, transgenic overexpression of the c-Myc oncogene in the liver activates hepatic glycolysis, leading to normalization of blood glucose levels.(40) Interestingly, c-Myc overexpression is a common feature of human and mouse hepatocellular neoplasms.(41,42) Hence, in our experiment, chronic hyperglycemia in the insulin-deficient mice might have provoked adaptive, insulin-independent glycolysis in the hepatic tumor cells.

Intriguingly, human hepatocellular carcinomas are also known to exhibit fatty change, especially in their early stages. The frequency is as high as 42.4% for those sized 1.1–1.5 cm in diameter and significantly declines for those either smaller or larger than that range.(43) It has been postulated that medium size is associated with an insufficient blood supply due to limited neovascularization, resulting in fatty degenerative damage.(43)

Shortage in serum adiponectin, a representative adipokine secreted by fat tissue, was apparent in insulin-deficient mice and this has been linked with human obesity, DM, and cancer risk.(34) Thus, it is tempting to assume that the low adiponectin contributes to the pronounced liver tumor growth in insulin-deficient mice. Although any definite mechanisms have yet to be clarified, it is postulated that anticancer effects of adiponectin are exerted through its antidiabetic, anti-inflammatory, and anti-angiogenic features.(44,45) On the other hand, even though another representative adipokine, leptin, is a suspected promoter of carcinogenesis,(35) the insulin-deficient mice were virtually null for leptin. However, some negative effects of leptin on cancer have also been reported(35) and no detailed study on leptin and liver cancer has been available. We therefore do not exclude the possibility that the leptin-null status in insulin-deficient mice played some role in the promotion of hepatocarcinogenesis.

The mean epididymal fat weight of insulin-deficient mice was only one-eighth of that for control mice. It is of note that genetically engineered fatless mice deficient in adipokines (including adiponectin and leptin) are diabetic with compensatory hyperinsulinemia and prone to tumor development due to unknown mechanisms.(44,46,47) Thus, apart from the insulin status, the insulin-deficient mice share some characteristics with the fatless mice, suggesting that these two models may also share some common mechanisms for cancer susceptibility. Although impaired adipokine secretion has not been well documented for lean individuals unlike obese individuals, excessive loss of fat may lead to reduced adipokine secretion as observed in insulin-deficient mice. This is highly relevant to the human case, because uncontrolled DM patients eventually experience a shortage of insulin secretion and suffer from severe weight loss. Consequently, in addition to the well-established obesity–cancer link, a possible mechanistic link between cancer and excessive or complete fat loss is an important issue to be pursued especially in view of reduced adipokine secretion.

We also quantified serum levels of Igf1 and its binding protein Igfbp3 as a proposed regulator of Igf1 bioavailability. Igf1 and insulin are derived from the same ancestral origin and share their receptors and functions to some degree.(32) Elevation in Igf1 is associated with human cancer risk as well as hyperinsulinemia.(32,33) Interestingly, the fatless mice were reported to show increased serum levels of Igf1.(46) However, we here found no significant difference in either Igf1 or Igfbp3 level between control and insulin-deficient mice.

Under the hypothesis that insulin promotes carcinogenesis, insulin therapy of DM patients could be an undesirable cancer risk.(48) Thus, we studied the effect of experimental insulin therapy on liver tumor-bearing, insulin-deficient mice. Relative to the saline-injected group, the insulin-injected group exhibited a higher apoptotic rate, but a similar Ki-67 labeling index in the tumors. Therefore, we could obtain no evidence that the insulin therapy promotes liver tumor growth. However, due to the difficulty in controlling blood glucose level, some of the mice were hypoglycemic and a long-term experiment was not feasible. Accordingly, we require further study with a more elaborate experimental design for insulin supply to substantiate the present results.

Finally, it may be worth mentioning that the mean normal liver volume of the insulin-deficient mice was larger than that of the control mice by 10%. Although the difference was marginally insignificant, this could be relevant to characteristic hepatomegaly reported for uncontrolled human diabetic patients.(49) Thus, the Akita mouse model might also be useful for studies on diabetes-related non-neoplastic alterations of the liver.

In summary, while our experimental results do not necessarily exclude compensatory hyperinsulinemia associated with type 2 DM to be a liver cancer risk, the promotion of hepatocellular tumor development observed in the present study is definitely insulin independent. This warrants further investigations on the insulin-independent mechanisms in order to fully understand the relationship between DM and liver cancer.


This work was supported in part by a Grant-in-Aid for scientific research from the Japan Society for the Promotion of Science, a President’s discretionary grant from Kochi University, and a grant from the Smoking Research Foundation.