Research ArticleLiver Cancer

S-Nitrosylation from GSNOR Deficiency Impairs DNA Repair and Promotes Hepatocarcinogenesis

See allHide authors and affiliations

Science Translational Medicine  17 Feb 2010:
Vol. 2, Issue 19, pp. 19ra13
DOI: 10.1126/scitranslmed.3000328


Human hepatocellular carcinoma (HCC) is associated with elevated expression of inducible nitric oxide synthase (iNOS), but the role of nitric oxide in the pathogenesis of HCC remains unknown. We found that the abundance and activity of S-nitrosoglutathione reductase (GSNOR), a protein critical for control of protein S-nitrosylation, were significantly decreased in ~50% of patients with HCC. GSNOR-deficient mice were very susceptible to spontaneous and carcinogen-induced HCC. During inflammatory responses, the livers of GSNOR-deficient mice exhibited substantial S-nitrosylation and proteasomal degradation of the key DNA repair protein O6-alkylguanine-DNA alkyltransferase. As a result, repair of carcinogenic O6-alkylguanines in GSNOR-deficient mice was significantly impaired. Predisposition to HCC, S-nitrosylation and depletion of alkylguanine-DNA alkyltransferase, and accumulation of O6-alkylguanines were all abolished in mice deficient in both GSNOR and iNOS. Thus, our data suggest that GSNOR deficiency, through dysregulated S-nitrosylation, may promote HCC, possibly by inactivating a DNA repair system.


Hepatocellular carcinoma (HCC) is the third leading cause of cancer deaths worldwide (1), and HCC incidence and mortality in the United States is rapidly increasing (2). The etiology of HCC is well established, with most HCC attributable to chronic hepatitis from hepatitis B or C virus infection (3). Nevertheless, the molecular mechanisms through which risk factors contribute to hepatocarcinogenesis remain, for the most part, poorly understood (4).

Inducible nitric oxide synthase (iNOS), the enzyme responsible for the high production of nitric oxide in the innate immune response and inflammation, is often increased (both mRNA and protein) in the hepatocytes of patients with chronic hepatitis B or C virus infection (57), hemochromatosis (8), and alcoholic cirrhosis (9), all of which predispose to HCC. Furthermore, iNOS is expressed in the hepatocytes within HCCs (7, 10), and HCC patients exhibit elevated concentrations of plasma nitrite or nitrate (11, 12). Nevertheless, studies on iNOS−/− mice, in spontaneous and fibrosis-associated models of HCC, revealed little effect of iNOS-derived nitric oxide on hepatocarcinogenesis (13). The amount of nitric oxide bioactivity, however, is controlled not only by nitric oxide synthases but also by enzymatic degradation (1416), and defective degradation can result in excessive amount of nitric oxide bioactivity in vivo (14). Whether nitric oxide plays a role in hepatocarcinogenesis remains unclear.

S-Nitrosylation is a major mechanism through which nitric oxide modifies the functions of proteins to exert control over biological processes (17). Protein S-nitrosylation is thus a potential modulator of cellular processes important for tumorigenesis, including inhibition or induction of apoptotic cell death and inhibition of DNA repair (17, 18). The DNA repair enzyme O6-alkylguanine-DNA alkyltransferase (AGT) repairs mutagenic and cytotoxic O6-alkylguanines, which can be mispaired by DNA polymerases to thymine during DNA replication, causing G:C to A:T transition (19). O6-alkylguanines are produced by alkylating N-nitroso compounds that are present in the environment and are formed endogenously through either NOS-dependent or NOS-independent pathways (1921). Mice deficient in AGT are more susceptible to HCC induced by dimethylnitrosamine (22), whereas overexpression of AGT in transgenic mice reduces both diethylnitrosamine (DEN)–induced and spontaneous HCC (23, 24), indicating a protective role of AGT against HCC. AGT can be inactivated by S-nitroso-N-acetylpenicillamine and S-nitrosoglutathione (GSNO) through S-nitrosylation of the cysteine in the enzyme active site in vitro (18). Nevertheless, the precise role of protein S-nitrosylation in the development of HCC is not clear.

GSNO reductase (GSNOR; also known as alcohol dehydrogenase class III), a ubiquitous, phylogenetically conserved enzyme, is the primary means of the cell for degrading the main nonprotein S-nitrosothiol (SNO), GSNO (14, 15). GSNO is in equilibrium with protein SNOs in cells, and GSNOR controls cellular concentrations of protein SNOs (1416, 25). Mice deficient in GSNOR exhibit large increases in protein S-nitrosylation and tissue injury after iNOS induction; the protective function of GSNOR against nitrosative stress is particularly prominent in the liver (14). The human GSNOR gene (ADH5) is located at approximately 4q23, a region in which chromosomal deletion occurs most frequently in HCC (2629). Furthermore, deletions in 4q23 occur frequently in cirrhotic and dysplastic hepatocytes, the precursor cells for HCC (26, 30). The gene or genes potentially important to HCC in the region remain to be identified.

During our study of GSNOR−/− mice, we noticed a high incidence of spontaneous liver tumors. Here, we have established that the protein amount and activity of GSNOR is frequently deficient in human HCC. We then used GSNOR−/− mice and found that GSNOR is important for protection against both spontaneous and DEN-induced HCC and investigated possible molecular mechanisms.


Frequent deficiency of GSNOR in human HCC

To investigate the potential deficiency of GSNOR in human HCC, we first measured GSNOR enzymatic activity in pairs of HCC and associated noncancerous liver tissues from 24 HCC patients (Fig. 1 and fig. S1). Whereas GSNOR activity in noncancerous liver with or without cirrhosis was similar (fig. S1A), the activity was significantly decreased in HCC, with reduction of GSNOR activity in 50% of the HCCs (Fig. 1, A and B). GSNOR activity was reduced by at least 50% in 10 (P1 to P10) of the 24 HCCs relative to the paired noncancerous tissues, and the activity in 5 (P1 to P5) of the HCCs was decreased by 80 to 90% (Fig. 1A and fig. S1B). In two additional patients (P19 and P21), GSNOR activity was low in both HCC and noncancerous tissues (Fig. 1A). For the other 12 patients, GSNOR activity in HCC appeared to be normal because it did not significantly differ from the activity in associated noncancerous liver or from the group mean of the noncancerous livers (Fig. 1, A and B). Thus, GSNOR activity was low in 50% of the HCC patients. To investigate whether deficiency of GSNOR activity in HCC results from reductions in GSNOR protein, we conducted quantitative immunoblot analysis of GSNOR (Fig. 1C). In patients who had lower GSNOR enzymatic activity in HCC than in normal tissue, the amount of GSNOR protein was also lower in HCC than in noncancerous tissue (Fig. 1C), and the decrease in GSNOR activity in HCC was correlated to the decrease in GSNOR protein amount (fig. S1C). Thus, both GSNOR enzymatic activity and amount of GSNOR protein are frequently decreased in human HCC; the reduced amount of protein is likely a main cause of the decrease in enzymatic activity.

Fig. 1.

GSNOR is frequently deficient in human HCC. (A) GSNOR enzymatic activity in HCC (red) and associated noncancerous liver (green) tissues from 24 patients. P, patient. Data are from two to four independent experiments, with the mean indicated by the bar. (B) The mean GSNOR enzymatic activity in the HCCs is significantly lower than that in the noncancerous liver tissues in the 24 patients (P = 0.01, Wilcoxon rank-sum test). (C) Western blot analysis of GSNOR protein in 14 of the 24 patients. NT, nontumor liver tissue; HCC, cancer tissue.

Predisposition of GSNOR−/− mice to spontaneous HCC

To study the role of GSNOR in tumorigenesis, we macroscopically and histologically analyzed age-matched wild-type C57BL/6 (n = 53) and congenic GSNOR−/− (n = 54) mice for spontaneous tumors (Fig. 2 and fig. S2). GSNOR−/− mice, on a C57BL/6 genetic background known to be resistant to spontaneous hepatotumorigenesis (31), frequently developed hepatocellular adenomas (HCAs) and HCCs (Fig. 2) and, in one instance, hepatoblastoma (fig. S2). Hepatocellular tumors between 5 and 20 mm in diameter started to appear in GSNOR−/− mice at 1.5 years of age, and the tumor-free survival of GSNOR−/− mice was significantly decreased (Fig. 2C). GSNOR−/− mice frequently developed multiple hepatocellular tumors, often consisting of both HCA and HCC (Fig. 2 and fig. S2, A and B). The incidence of HCA and HCC in GSNOR−/− mice was, respectively, 4 and 10 times that in wild-type controls (Fig. 2D). The incidence of hepatocellular tumors was increased in both male (Fig. 2E) and female (Fig. 2F) GSNOR−/− mice, with a greater increase in males. The incidence of all types of hepatocellular tumors and of HCC in GSNOR−/− males was about 50% and 30%, respectively. Thus, GSNOR−/− mice are predisposed to develop spontaneous HCA and HCC.

Fig. 2.

GSNOR−/− mice are predisposed to spontaneous hepatocellular tumors. (A) Liver of a GSNOR−/− mouse with two tumors (arrows). (B) H&E-stained section of HCC from a GSNOR−/− mouse. Scale bar, 100 μm. (C) Hepatocellular (HC) tumor-free survival curves from Kaplan-Meier analysis. P = 0.007, log-rank test, wild-type (WT) (n = 53) versus GSNOR−/− (n = 54). (D to F) Incidence of HCC, HCA, and all hepatocellular tumors (All) in all the mice (D), males (E) (26 wild type, 29 GSNOR−/−), and females (F) (27 wild type, 25 GSNOR−/−). Tumor incidence is shown as a percentage of mice with tumor. Significant differences between wild-type and age-matched GSNOR−/− mice are indicated by one (P < 0.05) or two (P < 0.005) asterisks.

The incidence of other spontaneous tumors, including those of spleen, kidney, heart, and lung, was not significantly increased in GSNOR−/− mice (fig. S2C). For instance, the incidence of spontaneous lymphoma, a common neoplasm in aged mice, was 20% in GSNOR−/− (n = 54) and 24% in wild-type C57BL/6 (n = 45) mice. Thus, the increase in spontaneous tumorigenesis in GSNOR−/− mice in the experimental setting appears to be specific to the liver.

Spontaneous hepatotumorigenesis in a number of transgenic mouse models is associated with chronic liver injury and increased hepatocyte turnover (3237). However, serum concentrations of alanine aminotransferase, a marker of liver injury, did not differ between tumor-free GSNOR−/− mice and wild-type controls (fig. S3A) (14). The histology of the noncancerous liver tissue of GSNOR−/− mice was indistinguishable from that of age-matched wild-type controls and showed no sign of increased inflammation (fig. S3, B and C). In addition, immunohistochemical staining for Ki67, a marker of proliferating cells, showed no difference between the livers of GSNOR−/− mice and wild-type controls before HCC development (fig. S4). Thus, GSNOR−/− mice do not appear to suffer chronic liver injury and compensatory regeneration.

Increase in spontaneous hepatocarcinogenesis in GSNOR−/− mice is abolished by genetic deletion of iNOS

To determine whether iNOS activity contributes to spontaneous hepatocarcinogenesis in GSNOR−/− mice, we studied the incidence of hepatocellular tumors in GSNOR−/−iNOS−/− double-knockout mice. GSNOR−/−iNOS−/− mice, congenic to C57BL/6, develop and reproduce normally. Unlike the high incidence of HCA and HCC in GSNOR−/− males, the age-matched GSNOR−/−iNOS−/− males showed a tumor incidence identical to that of wild-type controls (Fig. 3). Thus, nitric oxide bioactivity, possibly SNO from iNOS, predisposes liver to spontaneous HCC in the absence of GSNOR, and protection against the hepatocarcinogenic activity of iNOS requires a physiological function of GSNOR.

Fig. 3.

Spontaneous hepatocellular tumors in the GSNOR−/− background are decreased by deletion of iNOS. GSNOR−/−iNOS−/− (n = 28), GSNOR−/− (n = 29), and wild-type (n = 26) males were analyzed histopathologically. Tumor incidence is shown by a percentage of mice with tumor. Incidence of all hepatocellular tumors (P = 0.007) and of HCC alone (P = 0.013) in GSNOR−/−iNOS−/− mice was significantly lower than in GSNOR−/− mice.

GSNOR−/− mice are more susceptible to chemical hepatocarcinogenesis

To further study the role of GSNOR in hepatocarcinogenesis, we used GSNOR−/− mice in a DEN-induced HCC model (38). DEN, one of a group of alkylating N-nitroso compounds both present in the environment and formed endogenously (20, 21, 39), is a genotoxic compound that causes tumorigenesis predominantly in the liver (39). We treated 15-day-old male mice with a single intraperitoneal injection of DEN (5 μg/g) and analyzed DEN-induced tumor formation at the age of 10.5 months, an age at which spontaneous liver tumors have not occurred. The DEN-treated GSNOR−/− mice developed more than 5 times as many liver tumors as the DEN-treated wild-type controls (Fig. 4, A to C). Many of the tumors in a DEN-treated GSNOR−/− mouse were much larger than the ones in a wild-type control (Fig. 4, A and B), and the maximal tumor diameters in GSNOR−/− mice were more than 4 times those of wild-type controls (Fig. 4D). The increase in tumor multiplicity and maximal size in GSNOR−/− mice was abolished in GSNOR−/−iNOS−/− mice (Fig. 4, C and D). Thus, loss of GSNOR, possibly through inability to control iNOS-derived SNO, significantly promotes DEN-induced hepatocarcinogenesis.

Fig. 4.

DEN-induced tumor development is increased in GSNOR−/− mice. (A and B) Representative livers of male wild-type and GSNOR−/− mice 10 months after DEN (5 μg/g) injection. (C and D) Number of tumors (>1 mm) per mouse (C) and maximal tumor diameters (mean ± SE) (D) in DEN-treated wild-type (n = 13), GSNOR−/− (KO) (n = 10), and GSNOR−/−iNOS−/− (DKO) (n = 14) mice. GSNOR−/− mice develop significantly more tumors than wild-type (P < 0.001) and GSNOR−/−iNOS−/− (P < 0.001) mice; the maximal tumor size in GSNOR−/− is significantly bigger than that in wild-type (P < 0.001) and GSNOR−/−iNOS−/− (P = 0.003) mice.

DEN-challenged GSNOR−/− mice exhibit marked decrease in AGT

GSNOR could potentially promote breakdown of mutagenic N-nitroso compounds through glutathione-mediated denitrosation. However, DEN metabolism does not involve glutathione-mediated denitrosation (39); consequently, we did not expect GSNOR to promote direct denitrosation and inactivation of DEN. Although GSNOR could potentially protect many cellular processes, one possible candidate is DNA repair. DEN, after being activated in hepatocytes, causes a potent mutagenic and carcinogenic DNA lesion, O6-ethylguanine (39). DEN-induced hepatocarcinogenesis is inhibited by overexpression of the DNA repair enzyme AGT (23). To investigate a possible effect of GSNOR on AGT, we measured AGT activity in mouse livers. The amount of liver AGT activity was not different between resting GSNOR−/− and wild-type mice (fig. S5). However, 6 days after a single intraperitoneal injection of DEN, the liver AGT activity in GSNOR−/− mice was reduced by ~80% relative to the wild-type control (Fig. 5A). Liver lactate dehydrogenase (LDH) activity, on the other hand, did not differ among the DEN-challenged mice (Fig. 5B). Thus, in response to DEN challenge, liver AGT activity was selectively decreased in GSNOR−/− mice. DEN causes liver inflammation (40), and one major inflammatory mediator is iNOS. The amount of liver AGT activity was not lower in DEN-challenged GSNOR−/−iNOS−/− mice than in the wild-type controls (Fig. 5A). Thus, AGT activity in the DEN-challenged GSNOR−/− mouse may be decreased by SNO from iNOS, and GSNOR is likely required to protect AGT from nitrosative inactivation.

Fig. 5.

AGT is depleted in livers of DEN-challenged GSNOR−/− mice. (A) AGT activity in liver lysates of mice 6 days after DEN (37.5 μg/g) injection. Data are the mean (± SE) from four wild-type, five GSNOR−/− (KO), and three GSNOR−/−iNOS−/− (DKO) mice. The AGT activity in GSNOR−/− mice is significantly lower than that in wild-type (P < 0.004) and GSNOR−/−iNOS−/− (P < 0.001) mice. (B) LDH activity in liver lysates of the DEN-challenged mice described in (A). Data are the mean (± SD) from four wild-type, three GSNOR−/−, and three GSNOR−/−iNOS−/− mice. (C) Immunoblot of AGT, GSNOR, and β-actin in livers of wild-type, GSNOR−/−, and GSNOR−/−iNOS−/− mice before or 6 days after DEN injection. The amount of liver lysate in the last lane, 7.5 μg, is 30% of that in the other lanes. (D) Immunoblot of AGT and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) in livers of mice 4 days after treatment with DEN or DEN followed by the proteasome inhibitor MG262.

We carried out immunoblot analysis with an antibody to mouse AGT and found that, after DEN challenge, the concentration of AGT protein in the liver of GSNOR−/− mice was much lower than that in wild-type and GSNOR−/−iNOS−/− mice (Fig. 5C). The AGT protein concentration did not differ among the mice that did not receive DEN. These data indicate that the SNO-dependent decrease in AGT activity in DEN-challenged GSNOR−/− mice results from reduction of AGT protein.

Inhibition of DNA repair in DEN-challenged GSNOR−/− mice

To determine whether nitrosative inactivation of AGT is associated with impaired repair of carcinogenic O6-alkylguanines, we analyzed genomic DNA from livers of DEN-challenged mice by immuno-slot blot with a monoclonal antibody against O6-ethyldeoxyguanosine (41). Although no O6-ethyldeoxyguanosine was detected without DEN challenge, 2 days after DEN injection, substantial amounts of O6-ethyldeoxyguanosine were detected and the concentrations of O6-ethyldeoxyguanosine were equivalent in wild-type, GSNOR−/−, and GSNOR−/−iNOS−/− mice (Fig. 6). Whereas O6-ethyldeoxyguanosine, as expected, was mostly repaired by day 6 after DEN injection in wild-type mice, the amount of O6-ethyldeoxyguanosine at this time was higher in GSNOR−/− mice. Furthermore, repair of O6-ethyldeoxyguanosine was restored by day 6 in GSNOR−/−iNOS−/− mice. Thus, repair of carcinogenic O6-alkylguanine is impaired by nitrosative stress in GSNOR−/− mice. In addition, the concentration of O2-ethyldeoxythymidine, the long-lasting DNA lesion from DEN that is not repaired by AGT (39), was comparable among DEN-treated wild-type, GSNOR−/−, and GSNOR−/−iNOS−/− mice (Fig. 6), which suggests that the effect of GSNOR deficiency on DNA lesion appears to be specific to O6-alkylguanine.

Fig. 6.

Repair of O6-ethyldeoxyguanosine in livers of GSNOR−/− mice is impaired. Immuno-slot blot of genomic DNA from livers of wild-type, GSNOR−/− (KO), and GSNOR−/−iNOS−/− (DKO) mice before (−DEN) or 2 and 6 days after DEN injection. The DNA (1 μg per slot) is probed with monoclonal antibodies against O6-ethyldeoxyguanosine (O6-ethyl-dG) and O2-ethyldeoxythymidine (O2-ethyl-dT), respectively.

Depletion of AGT through proteasomal degradation in GSNOR−/− mice during inflammatory responses

To examine protection of AGT from nitrosative inactivation by GSNOR as a general mechanism during inflammatory responses, we studied AGT in livers of mice challenged with an intraperitoneal injection of lipopolysaccharide (LPS), which causes systematic inflammation and iNOS expression in hepatocytes (14, 42). Whereas the LPS challenge had little effect on liver AGT protein abundance in wild-type mice, the treatment resulted in almost complete loss of AGT protein in GSNOR−/−mice (Fig. 7A). Loss of AGT after LPS challenge was prevented in GSNOR−/−iNOS−/− mice (Fig. 7A). Thus, GSNOR is required to protect AGT from nitrosative inactivation after LPS treatment. Furthermore, AGT in hepatocytes isolated from GSNOR−/− mice was more susceptible to nitrosative inactivation than that in wild-type hepatocytes (fig. S6), indicating protection of AGT by GSNOR in the cells autonomously. Thus, AGT in liver cells appears to be highly susceptible to nitrosative stress and dependent on GSNOR for protection.

Fig. 7.

S-Nitrosylation and inactivation of AGT in livers of GSNOR−/− mice. (A) Immunoblot of AGT and β-actin in livers of wild-type, GSNOR−/− (KO), and GSNOR−/−iNOS−/− (DKO) mice treated with no LPS (−LPS), LPS (+LPS), or LPS followed by the proteasome inhibitor MG262 (+LPS +MG262). (B) Detection of S-nitrosylated AGT (SNO AGT) by the biotin switch assay and AGT immunoblot. Samples were GSNOR−/−iNOS−/− liver lysate treated with no chemical (−), 300 μM cysteine (Cys), 300 μM S-nitroso-cysteine (SNO), 1 mM DTT, or 300 μM hydrogen peroxide (H2O2). Bottom row, immunoblot of total AGT in each sample. (C) Biotin switch analysis of endogenously S-nitrosylated AGT in livers of LPS-challenged mice. MG262 was injected into some of the mice to inhibit proteasomal degradation of AGT. Bottom row, total AGT.

Intact AGT protein is quite stable, but after inactivation, the protein is rapidly degraded through the ubiquitin-proteasome pathway (18, 43). To investigate the role of proteasomal degradation in the control of AGT protein abundance in GSNOR−/− mice during inflammatory responses, 2 days after LPS challenge, we gave mice an intraperitoneal injection of MG262, a specific inhibitor of the proteasome (44). Although treatment with MG262 for ~4 hours had little effect on liver AGT concentration in wild-type and GSNOR−/−iNOS−/− mice, the treatment in GSNOR−/− mice brought the AGT protein concentration close to that of resting mice (Fig. 7A). Similarly, in DEN-treated mice, MG262 treatment, equivalent to that described above, largely prevented the decrease in AGT in GSNOR−/− mice (Fig. 5D). In addition, after inhibition of the proteasome by MG262 treatment in DEN-challenged mice, we detected more polyubiquitinated AGT in GSNOR−/− than in wild-type and GSNOR−/−iNOS−/− mice (fig. S7). Thus, the SNO-dependent decrease in AGT during inflammatory responses in GSNOR−/− mice is brought about by proteasomal degradation of (presumably damaged) AGT.

S-Nitrosylation of AGT in liver of GSNOR−/− mice

AGT is susceptible to S-nitrosylation and inactivation by GSNO in vitro (18). We found that S-nitrosylated AGT formed by treatment with S-nitroso-cysteine can be specifically detected by the biotin switch method (45) in conjunction with an antibody to AGT (Fig. 7B). This SNO assay does not recognize AGT in liver lysate treated with other redox chemicals, including hydrogen peroxide (Fig. 7B). When proteasomal degradation of AGT was inhibited by MG262, we detected AGT S-nitrosylation that had been formed endogenously in the livers of LPS-challenged GSNOR−/− mice (Fig. 7C). The concentration of S-nitrosylated AGT was higher in GSNOR−/− than in wild-type control, and AGT SNO was abolished in GSNOR−/−iNOS−/− mice (Fig. 7C). AGT is thus a direct target of S-nitrosylation in vivo, and GSNOR is required to prevent abnormally elevated S-nitrosylation of AGT from iNOS during inflammatory responses.


GSNOR deficiency in human HCC

We found that the amount of GSNOR in HCC in ~50% of patients is decreased by 50 to 90%. This common deficiency of GSNOR is consistent with a recurrent chromosomal deletion in human HCC at 4q22-23, which contains the human GSNOR gene (2629). In addition to loss of one of the two alleles of the GSNOR gene, however, there may be other defects (genetic or epigenetic) that contribute to the 80 to 90% reduction in GSNOR activity in ~20% of HCCs. Normal GSNOR activity in HCCs in 50% of the patients suggests that hepatocyte dedifferentiation during HCC development is unlikely by itself to cause GSNOR deficiency. It remains to be investigated whether reduction of GSNOR activity in noncancerous liver, as in patients P19 and P21, might predispose to HCC. Because GSNOR prevents S-nitrosylation–mediated degradation of the key DNA repair protein AGT and protects against hepatocarcinogenesis in mice, we suggest that a deficiency of GSNOR with concurrent overexpression of iNOS in human liver (59) results in dysregulated S-nitrosylation that is likely to be a common contributing factor in human HCC development. Indeed, gene expression profiling of liver tissue adjacent to HCC showed that both GSNOR deficiency and iNOS overexpression are closely associated with a poor prognosis in HCC patients (46).

Protection of mice from HCC by GSNOR

A principal discovery of this study is that GSNOR, the key protein in the control of SNOs, is critically important for protection against hepatocarcinogenesis in animals. GSNOR−/− mice were more susceptible to both spontaneous and carcinogen-induced HCC. Predisposition of GSNOR−/− mice to HCC was abolished by genetic deletion of iNOS. Thus, GSNOR protects mice from HCC, most likely through its physiological action on SNOs. This conclusion is further supported by the observation that GSNOR is required to prevent iNOS-dependent S-nitrosylation and depletion of the key DNA repair enzyme AGT in inflammatory responses.

Before the appearance of spontaneous HCC, the mice that are derived from targeted deletion of the genes mdr2 (35), Acox1 (33), Pten (34), Nrf1 (36), Sod1 (32), and Nemo (37) sustain chronic liver injury and consequent repeated rounds of hepatocyte death and regeneration, a process in which the repeated rounds of DNA replication and inflammation are thought to increase DNA mutation and hepatocarcinogenesis (3237). Mice transgenic for the core gene of hepatitis C virus, on the other hand, develop HCC spontaneously at old age in the absence of necroinflammation (47). Similarly, GSNOR−/− mice do not appear to suffer chronic liver injury. Thus, spontaneous hepatocarcinogenesis in GSNOR−/− mice does not result from chronic liver injury and the consequent hepatocyte turnover that is common in most other transgenetic models of spontaneous HCC.

Our study has also revealed a pro-hepatocarcinogenic effect of iNOS in the absence of GSNOR. Inducible NOS is constitutively expressed in ileal epithelium of normal mice (48) and helps to prevent opportunistic infection by commensal gastrointestinal microorganisms in liver and other organs in mice reared under specific pathogen-free (not germ-free) conditions (49). Thus, even in the absence of chronic inflammation, liver cells in our mice may be exposed to reactive nitrogen species from iNOS that is constitutively expressed or induced in response to commensal microorganisms. GSNOR deficiency results in altered response to iNOS activation, including nitrosative inactivation of AGT, and consequently makes the mice more susceptible to nitrosative stress. Inducible NOS is expressed in hepatocytes under conditions that predispose to HCC and in HCC itself in both animals and humans (59, 13, 50). Studies with iNOS−/− mice in spontaneous and fibrosis-associated models of HCC, nevertheless, revealed little effect of iNOS on hepatocarcinogenesis (13, 51). Our results suggest that the potential pro-hepatocarcinogenic activity of iNOS is normally prevented or masked by GSNOR. Induction of iNOS (52) and loss of chromosome 4q (53) occur in lung, breast, and other cancers. Thus, GSNOR deficiency and protein S-nitrosylation might also contribute to the development of other cancers.

Protection of AGT from S-nitrosylation and depletion by GSNOR

Our results suggest that one possible mechanism for the increased susceptibility to spontaneous and DEN-induced HCC in GSNOR−/− mice could be the S-nitrosylation and depletion of AGT during inflammatory responses. We showed previously that the concentration of liver protein SNOs, derived from iNOS activity, increases greatly during inflammation in GSNOR−/− mice (14). Our data now show that one of the proteins highly susceptible to S-nitrosylation by iNOS is AGT and that protection of AGT from S-nitrosylation requires GSNOR. In vitro recombinant human AGT is susceptible to S-nitrosylation at the enzyme active site Cys145 (18), where both the strong nucleophilicity of the cysteine sulfur and the close proximity of His146 (54) may promote S-nitrosylation (17). S-Nitrosylation of Cys145, which likely causes conformational changes in AGT (54), appears to be responsible for rapid proteasomal degradation of AGT when Chinese hamster ovary cells with transfected human AGT are treated with S-nitroso-N-acetylpenicillamine (18). S-Nitrosylation of AGT, possibly at the enzyme active site Cys149 with its nucleophilic sulfur and conserved His150 in close proximity, likely results in rapid degradation and loss of AGT in GSNOR−/− mice. Depletion of liver AGT in response to both LPS and DEN challenges in GSNOR−/− mice, but not in GSNOR−/−iNOS−/− mice, suggests that AGT is generally highly susceptible to nitrosative inactivation in inflammatory responses. Together, our results suggest that GSNOR protects AGT from hyper-S-nitrosylation and proteasomal degradation. In a few human HCC samples that are deficient in GSNOR activity, amounts of AGT protein are decreased (fig. S8), suggesting possible protection of AGT by GSNOR in humans.

Many DNA mutations affecting multiple cellular pathways are the hallmark of carcinogenesis (55, 56). We found that nitrosative inactivation of AGT, the protein critical for repair of carcinogenic O6-alkylguanines, impairs repair of O6-ethylguanine in the liver of GSNOR−/− mice, thus revealing a possible mechanism by which GSNOR deficiency could promote hepatocarcinogenesis. Unrepaired O6-ethylguanine is mispaired to thymine during DNA replication and causes G:C to A:T transition in the next round of DNA replication. G:C to A:T transition, which may arise from O6-alkylguanines or deamination of cytosine and 5-methylcytosine, is the most common mutation in cancer (55, 56). Accumulation of O6-alkylguanines from nitrosative inactivation of AGT likely increases DNA mutation and, through mutations of oncogenes and tumor suppressor genes, could promote initiation of DEN-induced hepatocarcinogenesis in GSNOR−/− mice. AGT overexpression in transgenic mice decreases the frequency of G:C to A:T mutations and reduces spontaneous HCC, which suggests that spontaneous O6-alkylguanine lesions can occur also (24). Although significant increase in spontaneous HCC is not detected in AGT-deficient mice up to ~10 months of age (22), the effect of AGT deficiency on spontaneous HCC in old mice close to 2 years old is unknown. Loss of AGT is associated with G:C to A:T mutation of key oncogenes and tumor suppressor genes in a number of human cancers (19). We thus suggest that nitrosative inactivation of AGT in inflammatory responses may allow mutagenesis from spontaneous O6-alkylguanine lesions and contribute significantly to spontaneous HCC in GSNOR−/− mice. In addition to nitrosative inactivation of AGT, iNOS-derived nitric oxide bioactivity may contribute to the formation of alkylating N-nitroso compounds and thus DNA mutation in hepatocarcinogenesis (20). Dysregulated S-nitrosylation from GSNOR deficiency may also accelerate HCC development, possibly through promoting the survival or growth of neoplastic cells (52, 57, 58).

Our results underscore the importance of protein S-nitrosylation in the pathogenesis of HCC. Thus, patients with GSNOR deficiency and concurrent iNOS overexpression in the liver may be at an increased risk of HCC, and inhibition of iNOS-derived S-nitrosylation in these patients may provide a therapeutic strategy to prevent HCC.

Materials and Methods

Human HCC

Human HCC and associated noncancerous liver tissues, collected during liver resections at the Stanford University and the University of Hong Kong, were frozen in liquid nitrogen within 0.5 hour after removal. All HCC samples were found by two pathologists to contain >80% tumor cells. Clinical data of HCC patients are provided in table S1. The study of the human samples was approved by the Internal Review Board of the University of California, San Francisco (UCSF).


GSNOR−/− mice (14), after backcrossing 10 times to C57BL/6, were bred with iNOS−/− mice (The Jackson Laboratory) to obtain GSNOR−/−iNOS−/− mice. All mice were maintained on normal mouse chow (5058 PicoLab Mouse Diet 20) in a specific pathogen-free facility at the UCSF. Analysis by polymerase chain reaction of liver samples from three GSNOR−/− mice found that they were not infected by Helicobacter (UC Davis Comparative Pathology Laboratory). The experimental protocol was approved by the Institutional Animal Care and Use Committee of UCSF.

Mouse hepatocarcinogenesis

Mice were either left untreated or given at day 15 a single intraperitoneal injection of DEN (5 μg/g) (38) to study spontaneous and chemical hepatocarcinogenesis, respectively. At necropsy, externally visible tumors (>1 mm in diameter) in the livers were counted and measured under stereomicroscopy. Samples were fixed with 4% paraformaldehyde in phosphate-buffered saline and embedded in paraffin. Tissue sections were stained with hematoxylin and eosin (H&E) and examined by a board-certified veterinary pathologist (M.A.H.) and a liver pathologist (S.K.). The hepatocellular tumor-free survival curve (with event defined as development of hepatocellular tumor) was calculated by the Kaplan-Meier method with JMP software, and early loss of tumor-free mice (because of skin lesions or other conditions) was included as censored cases in estimation of the curve.

GSNOR enzymatic activity

Liver homogenates were prepared on ice in a solution containing 20 mM tris-HCl (pH 8.0), 0.5 mM EDTA, 0.1% NP-40, 1 mM phenylmethylsulfonyl fluoride, and a protease inhibitor cocktail (Roche). The GSNOR activity was measured by GSNO-dependent NADH (reduced form of NAD+) consumption (3). Briefly, cell lysate (0.3 mg/ml) was incubated with 75 μM NADH in reaction buffer [20 mM tris-HCl (pH 8.0) and 0.5 mM EDTA] containing 0 or 100 μM GSNO at room temperature, and NADH fluorescence (absorption at 340 nm and emission at 455 nm) was measured over time to determine the initial rate of GSNO-dependent NADH consumption. In a number of samples, GSNOR activity was also measured by NADH-dependent GSNO consumption (15), with similar results.


Proteins in liver homogenates were separated by SDS–polyacrylamide gel electrophoresis, transferred to nitrocellulose membranes, and probed with rabbit antiserum to GSNOR (16), β-actin mouse monoclonal antibody (A-5441; Sigma), or goat antiserum to AGT (R&D Systems). GSNOR and β-actin in human samples were detected and quantified with infrared fluorescent secondary antibodies—a goat antibody to rabbit coupled to Alexa Fluor 680 (Molecular Probes) and a goat antibody to mouse coupled to IRDye 800 (Rockland Immunochemicals)—with an infrared fluorescence imaging system (Odyssey; LICOR Biosciences). AGT was detected with a donkey secondary antibody to goat coupled to horseradish peroxidase and SuperSignal West Femto Chemiluminescent Substrate (Pierce).

AGT and LDH activity

AGT activity was determined by measuring radioactivity of the [3H]methyl group transferred from DNA to AGT protein (18). The 3H-methylated DNA substrate, prepared by methylating calf thymus DNA with 0.05 mM N-[3H]methyl-N-nitrosourea (10 Ci/mmol; Amersham), was incubated with purified recombinant human AGT (positive control; Sigma), bovine serum albumin, or cell lysates (2 mg) in AGT buffer [50 mM tris (pH 7.6) and 1 mM EDTA] for 30 min at 37°C. The reaction was stopped by addition of a denaturing buffer [8 M urea, 20 mM tris (pH 7.6), and 60 mM NaCl]. After the remaining 3H-methylated DNA was completely hydrolyzed (in 1 M perchloric acid at 70°C for 2 hours) and removed, 3H radioactivity in [3H]methyl-labeled AGT protein was determined by scintillation counting. Background 3H radioactivity obtained with bovine serum albumin was subtracted to determine the concentration of AGT activity in each sample. AGT activity was normalized with protein concentration of the lysate. LDH activity was measured by pyruvate-dependent NADH consumption (59) in 100 μl of solution containing 50 mM tris-HCl (pH 8.0), 1 mM EDTA, 100 μM NADH, 1 mM sodium pyruvate, and 2 μg of liver lysates.

Immuno-slot blot

Genomic DNA was isolated from livers of DEN-challenged (37.5 μg/g, intraperitoneally, at day 15) mice, sonicated for 15 s with a Virsonic600 sonicator, denatured at 100°C for 10 min, and transferred at 1 μg per slot to nitrocellulose membrane with a Bio-Dot slot blotting apparatus (Bio-Rad). The DNA was probed first with either a rat monoclonal antibody (ER6, Axxora) against O6-ethyldeoxyguanosine or a mouse monoclonal antibody against O2-ethyldeoxythymidine (EM 4-1, Axxora) and then detected, respectively, with horseradish peroxidase–coupled goat antibody to rat or goat secondary antibody to mouse and Femto Chemiluminescent Substrate. Calf thymus DNA treated with ethyl nitrosourea [0.5 M in 50 mM tris (pH 7.5), 60 mM NaCl, and 0.5 mM EDTA at 37°C for 30 min] was used as a positive control of ethylated DNA lesions.

LPS treatment

LPS (Escherichia coli, serotype 026:B6; Sigma) at a dosage of 50 μg/g was injected intraperitoneally into adult female C57BL/6, GSNOR−/−, and GSNOR−/−iNOS−/− mice. A number of the mice at 40 hours after LPS were further injected intraperitoneally with 5 μg of MG262 (Calbiochem), and tissues were collected 4 hours later.

AGT S-nitrosylation

AGT S-nitrosylation was analyzed by the biotin switch method (45). Liver lysate was prepared in HEDN solution [250 mM HEPES (pH 7.5), 1 mM EDTA, 0.1 mM diethylenetriamine pentaacetic acid, and 0.1 mM neocuproine] supplemented with 1 mM phenylmethylsulfonyl fluoride. The lysate (12 mg of protein) was incubated in a reaction mixture (0.8 to 1.0 mg of protein per milliliter, 20 mM methyl methanethiosulfonate, and 2.5% SDS in HEDN) at 65°C for 30 min to block free thiols. After removal of methyl methanethiosulfonate by repeated (twice) acetone precipitation, the SNO–biotin switch reaction was carried out by incubating the protein sample (4 mg/ml) with 50 mM ascorbate, 1 mM N-[6-(biotinamido)hexyl]-3′-(2′-pyridyldithio)propionamide (biotin-HPDP; Pierce), and 1% SDS in HEDN at room temperature in the dark for 1 hour. The remaining biotin-HPDP was removed by repeated acetone precipitation, and the biotinylated proteins were purified with 250 μl of NeutAvidin agarose resins (Pierce), eluted with 100 mM 2-mercaptoethanol, and then subjected to Western blotting with an antibody to AGT to detect S-nitrosylated AGT. In addition, to confirm the specificity of the biotin switch assay, GSNOR−/−iNOS−/− liver lysate was incubated in HEDN solution with 300 μM S-nitroso-cysteine, 300 μM cysteine, 1 mM dithiothreitol (DTT), 300 μM hydrogen peroxide, or no additional chemicals at room temperature for 60 min; purified by acetone precipitation; and then subjected to the biotin switch assay.

Statistical analysis

GSNOR activity in HCC patients was analyzed by Wilcoxon rank-sum test with JMP IN statistical software. Kaplan-Meier tumor-free survival curves were analyzed by log-rank test. Tumor incidences were analyzed by both the χ2 test and the Fisher’s exact test of contingency tables, with similar results. All the other data were analyzed with the Student’s t test. Significance was set at P < 0.05.

Supplementary Material

Table S1. Clinical data of HCC patients.

Fig. S1. Human GSNOR in HCC and noncancerous liver tissues.

Fig. S2. Tumorigenesis in GSNOR−/− mice.

Fig. S3. No increase in liver injury or inflammation in aged tumor-free GSNOR−/− mice.

Fig. S4. No difference in proliferation index in livers of wild-type and GSNOR−/− mice before HCC development.

Fig. S5. AGT activity in liver of 3-week-old resting GSNOR−/− and wild-type mice.

Fig. S6. Protection of AGT from nitrosative inactivation by GSNOR autonomously in hepatocytes.

Fig. S7. Increase in AGT ubiquitination in GSNOR−/− mice.

Fig. S8. AGT in human HCCs.


  • Citation: W. Wei, B. Li, M. A. Hanes, S. Kakar, X. Chen, L. Liu, S-Nitrosylation from GSNOR Deficiency Impairs DNA Repair and Promotes Hepatocarcinogenesis. Sci. Transl. Med. 2, 19ra13 (2010).

References and Notes

  1. Acknowledgments: We thank S.-T. Cheung and S. T. Fan at the University of Hong Kong for providing the human liver samples; UCSF Liver Center/San Francisco Veterans Administration Medical Center pathology core, M. Foster, H. Willenbring, and M. Egeblad for technical assistance; and L. Coussens, J. Stamler, Z. Werb, and B. Yen for reading the manuscript and comments. Funding: Sandler Family Supporting Foundation, Stewart Trust Cancer Research Award, UCSF Liver Center grant P30 DK26743, UC Cancer Research Coordinating Committee Research Grant, and NIH grant CA122359 (L.L.). Author contributions: W.W. designed and performed the experiments and analyzed the data; B.L. performed the experiments of clinical samples and analyzed the data; M.A.H. and S.K. provided histopathologic interpretation of tissue specimens; X.C. provided clinical samples, data, and expertise; L.L. designed the research, analyzed the data, and wrote the paper. Competing interests: The authors declare no competing interests.
View Abstract

Navigate This Article