Evidence That Hydrogen Sulfide Is a Genotoxic Agent

Hydrogen sulfide (H₂S) produced by commensal sulfate-reducing bacteria, which are often members of normal colonic microbiota, represents an environmental insult to the intestinal epithelium potentially contributing to chronic intestinal disorders that are dependent on gene-environment interactions. For example, epidemiologic studies reveal either persistent sulfate-reducing bacteria colonization or H₂S in the gut or feces of patients suffering from ulcerative colitis and colorectal cancer. However, a mechanistic model that explains the connection between H₂S and ulcerative colitis or colorectal cancer development has not been completely formulated. In this study, we examined the chronic cytotoxicity of sulfide using a microplate assay and genotoxicity using the single-cell gel electrophoresis (SCGE; comet assay) in Chinese hamster ovary (CHO) and HT29-Cl.16E cells. Sulfide showed chronic cytotoxicity in CHO cells with a %C1/2 of 368.57 μmol/L. Sulfide was not genotoxic in the standard SCGE assay. However, in a modified SCGE assay in which DNA repair was inhibited, a marked genotoxic effect was observed. A sulfide concentration as low as 250 μmol/L (similar to that found in human colon) caused significant genomic DNA damage. The HT29-Cl.16E colonocyte cell line also exhibited increased genomic DNA damage as a function of Na₂S concentration when DNA repair was inhibited, although these cells were less sensitive to sulfide than CHO cells. These data indicate that given a predisposing genetic background that compromises DNA repair, H₂S may lead to genomic instability or the cumulative mutations found in adenomatous polyps leading to colorectal cancer. (Mol Cancer Res 2006;4(1):9–14)

Colorectal cancer is the third most frequent cancer worldwide, producing 945,000 new cases and being responsible for 492,000 deaths annually (1). Genetic and environmental factors play a significant role in the development of colorectal cancer (2-5). Although etiologically divided into sporadic (∼90% of the cases), hereditary (5-10%), and inflammatory bowel disease associated (2%; ref. 1), all colorectal cancers show multistep development with several mutations (3-5). Genomic instability, a characteristic feature of colorectal cancer, potentially contributes to the multistep acquisition of mutations found in these tumors (6, 7) and to cancer formation (6, 8). Chromosomal instability is found in >80% of sporadic colorectal cancer (9), whereas microsatellite instability is highly associated with hereditary nonpolyposis colorectal cancer (9, 10). Several mechanisms have been proposed in chromosomal instability formation, including aneuploidy, cell cycle checkpoint loss, break-induced replication, defects in double-strand repair-recombination machinery, mitotic nondisjunction, and oxidative stress from free radicals produced by intraluminal bacteria (11-14). However, none of these potential mechanisms have been proven satisfactory, leaving the nature of this fundamental process in colorectal tumorigenesis unclear.

Numerous clinical and epidemiologic studies have evaluated the relationship between sulfate-reducing bacteria (often members of the normal colonic microbiota) or hydrogen sulfide (H₂S) and inflammatory bowel disease (such as ulcerative colitis) or colorectal cancer (15-20). Potential mechanisms underlying these clinical observations have focused mainly on the possibility that sulfide, through its toxic properties, damages the epithelium (21-26), thereby inciting chronic inflammation. Few studies have examined the effect of H₂S on intestinal epithelial cell fate, especially in the context of its potential contribution to sporadic colorectal cancer (27). Christl et al. (23) observed a significant increase in the proliferation of cells from a colonic biopsy incubated for 4 hours with 1 mmol/L NaHS. Deplancke and Gaskins (24) showed that H₂S may perturb the balance between apoptosis, proliferation, and differentiation in intestinal epithelial cells. Leschelle et al. (28) observed that sulfide decreased HT29 cell growth without increasing apoptosis or necrosis. In contrast to cell function studies, the toxic properties of H₂S have been extensively studied (29). Nevertheless, little is known about H₂S-induced DNA damage, its function as a mutagen or carcinogen, or its involvement in chromosomal instability formation. In the present study, using the alkaline single-cell gel electrophoresis (SCGE) assay or comet assay, we observed that H₂S is capable of generating DNA damage that may be, in part, responsible for the generation of the genomic instability and the cumulative mutations observed in colorectal cancer.

The adverse chronic cellular effect of H₂S was studied using a microplate method that measured the reduction in cell density as a function of the Na₂S concentration over a 72-hour period. In this assay, a decreased cell density could be due to a disruption in the cell cycle, to growth retardation, or to cell toxicity (30, 31). The data from repeated cytotoxicity experiments were averaged and plotted (Fig. 1), and regression analysis was used to calculate the %C1/2 value for Na₂S (which is the Na₂S concentration that causes a 50% reduction of the cell density compared with the negative control; ref. 31). Sodium sulfide had a %C1/2 value of 368.6 μmol/L, indicating that Na₂S is a moderate cytotoxic agent when compared with a previously published cytotoxicity library (31, 32). Nevertheless, Na₂S was more cytotoxic than the positive controls ethylmethane sulfonate (%C1/2 = 4.19 mmol/L) and potassium bromate (%C1/2 = 964 μmol/L; ref. 32). Because sulfide was cytotoxic at a concentration range similar to those found in the mouse intestine (0.2-1 mmol/L; ref. 33) and human feces (0.3-3.4 mmol/L H₂S; refs. 34-36), it is not clear how cells can survive long periods of exposure to H₂S. Different cell-specific detoxification mechanisms (37) or constant cell removal may be the strategy that the intestinal epithelium follows to maintain its integrity. It is also noticeable that the %C1/2 value is much higher than the 50 μmol/L previously reported to inhibit mitachondrial oxidative phosphorylation (38). In fact, in our study, at 50 μmol/L, no growth difference was observed compared with control cells.

The acute cytotoxic and the genotoxic characteristics of Na₂S were evaluated with Chinese hamster ovary (CHO) cells using the SCGE microplate assay. Significant changes in the average median tail moment were not observed in the concentration range of 25 to 5,000 μmol/L compared with the negative control (Fig. 2). Acute cytotoxicity after 4 hours of treatment was minimal for concentrations similar to those reported for the human colon and feces but increased drastically at ≥7,500 μmol/L concentrations, where only a few cells remained attached to the well after 4 hours of treatment.

A modified SCGE assay, where DNA repair was inhibited using hydroxyurea and 1-β-d-arabinofuranosylcytosine (AraC), was used to evaluate whether the previous negative observation was due to the lack of sulfide genotoxicity or was a consequence of active DNA repair. Hydroxyurea depletes the pools of deoxynucleotide triphosphate (39) by inhibiting ribonucleotide reductase, the enzyme responsible for deoxyribonucleotide biosynthesis (40). AraCTP, which is the major intracellular form of AraC, is incorporated into cellular DNA blocking DNA elongation by inhibiting DNA polymerase (39, 41). Figure 3 illustrates a concentration response increase of the average median tail moment values as a function of Na₂S concentration in the hydroxyurea and AraC pretreated cells. A significant difference (F_5,34 = 91.71; P < 0.001) in the average median SCGE tail moments was observed compared with the negative control. Sulfide concentrations as low as 250 μmol/L, which are commonly found in human feces (34-36), were genotoxic in mammalian cells (P < 0.001), indicating its possible role as an environmental insult in the development of mutations and eventually colorectal cancer. Examples of ethidium bromide–stained nuclei showing the concentration-dependent genotoxic effect of sulfide are presented in Fig. 4.

To evaluate the effect of Na₂S on cultured human colonocytes, we used the Cl.16E subclone of the human colonic cancer cell line HT29 (42) in the modified SCGE assay. The preliminary results are illustrated in Fig. 5. Compared with the transgenic CHO cells that were specifically generated for genotoxicity assays, the HT29-Cl.16E cells were less sensitive to Na₂S. Treatment with Na₂S alone or treatment with hydroxyurea and AraC without Na₂S did not induce a significant increase in genotoxic damage. However, Na₂S in a range from 500 to 2,000 μmol/L induced a positive concentration response with 2,000 μmol/L Na₂S significantly higher than the hydroxyurea + AraC negative control (F_5,14 = 13.63; P < 0.001). Under all treatment groups, the viability of the HT29-Cl.16E cells was >95%. Additional studies with a range of colonocyte cell lines or, ideally, primary colonocytes are needed to determine the extent to which colonocytes are resistant to genotoxic properties of sulfide.

To our knowledge, only one previous study investigated the mutagenic potential of H₂S (43). Using the Salmonella typhimurium mutagenicity assay with and without Aroclor-induced hamster and rat liver S9 fractions, Hughes et al. found that H₂S vapor (17-1750 g/plate) was not mutagenic in S. typhimurium strains TA97, TA98, or TA100 with and without metabolic activation (43). Despite that, Berglin and Carlsson showed that H₂S gas potentiated the mutagenicity of hydrogen peroxide in S. typhimurium strain TA102 (44), a strain specifically sensitive to DNA damage mediated by oxidative stress. Sulfide facilitated the formation of iron sulfide, which converts hydrogen peroxide to hydroxyl radicals more efficiently than ferrous iron. Consistently, Moore et al. (45) described the formation of sulfur-centered radicals as a consequence of superoxide production by Enterococcus faecalis that may be responsible in part for the action of sulfide in vivo.

c-Jun NH₂-terminal kinase is one of the mitogen-activated protein kinases that phosphorylate and activate the transcriptional activities of c-Jun and ATF2 in response to DNA damage (46, 47). Activation of c-Jun and ATF2 leads to a large-scale coordinated gene expression of several genes involved in DNA repair (48). Deplancke and Gaskins (24) observed an increase in steady-state c-Jun mRNA expression after exposing IEC-18 cells to 0.05 to 5 mmol/L NaHS. This observation was accompanied by a 3-fold increase in the DNA damage–inducible protein (Gadd45), consistent with the present observation that sulfide is capable of generating DNA damage.

The present data show that sulfide is genotoxic at concentrations as low as 250 μmol/L similar to that found in rat colon and human feces. DNA strand breaks have been directly associated with the induction of mutation promotion of induced carcinogenesis (49-51); therefore, H₂S may be responsible for some of the mutations needed in the development of colorectal cancer. Furthermore, H₂S may also be at least in part responsible for the genomic instability observed in colorectal cancer that potentially contributes to the multistep acquisition of mutations found in these tumors.

Colorectal cancer is a disease with both genetic and environmental determinants (3-5, 52). Polymorphisms in DNA repair genes have been associated with an increase in cancer risk (53, 54). Clinical studies have investigated the association in polymorphisms in xeroderma pigmentosum [XPD (or ERCC2), XPF, XPG, and ERCC1] and X-ray repair cross-complementing groups 1 and 3 (XRCC1 and XRCC3) genes and colorectal cancer risk (55-58). These studies show that polymorphisms in XRCC1 and XRCC3 genes are linked with an increase in colorectal cancer risk, although some results are ambiguous and depend on environmental and racial differences (55, 56, 58). In this study, the DNA damage generated by sulfide was only observed when DNA repair was inhibited, indicating that CHO cells actively repaired the damaged DNA. Individuals with a reduced ability to repair damaged DNA due to allelic polymorphisms in genes involved in DNA repair or harboring germ-line mutations in mismatch repair genes (Lynch syndrome; refs. 4, 59) or in the MYH (MUTYH) gene (59, 60) would be more sensitive to the environmental insult that H₂S represents. Even more, if alleles that favor colonic colonization by sulfate-reducing bacteria are also present, these subjects would be more susceptible to acquire different mutations and eventually develop colorectal cancer.

In summary, we observed that sulfide, which is produced by commensal sulfate-reducing bacteria in the large intestine, is capable of generating genomic DNA damage. Given a predisposing genetic background, this may lead to genomic instability or the cumulative mutations found in adenomatous polyps and colorectal cancer.

General laboratory reagents were purchased from Fisher (Itasca, IL) or Sigma (St. Louis, MO). Na₂S·9H₂O was obtained from Sigma. Media supplies and fetal bovine serum (FBS) were purchased from Hyclone Laboratories (Logan, UT). Clone 11-4-8 of CHO cell line AS52 was maintained in Ham's F12 medium containing 5% FBS, 1% antibiotics (100 units/mL sodium penicillin G, 100 μg/mL streptomycin sulfate, 0.25 μg/mL amphotericin B, 0.85% saline), and 1% glutamine at 37°C in a humidified atmosphere of 5% CO₂ (61). HT29-Cl.16E cells (ref. 42; kind gift from C.L. Laboisse, Université de Nantes, France) were maintained in DMEM supplemented with 10% heat-inactivated FBS and 0.5% antibiotics (50 units/mL penicillin G, 50 mg/mL streptomycin sulfate, 0.25 mg/mL amphotericin B, 0.85% saline) at 37°C in a humidified atmosphere of 5% CO₂.

Chronic cytotoxicity to mammalian cells was measured as described previously (30-32). Flat-bottomed, tissue culture 96-well microplates were used; eight replicate wells were prepared for each concentration of Na₂S·9H₂O. Eight wells were reserved for the blank control consisting of 200 μL F12 + 5% FBS. The negative control consisted of eight wells containing 100 μL of a titered CHO cell suspension (3 × 10⁴ cells/mL) plus 100 μL F12 + FBS. The wells for the remaining columns contained 3,000 CHO cells, F12 + FBS, and a known concentration of Na₂S for a total of 200 μL. To prevent crossover contamination between wells due to volatilization of the test agent, a sheet of sterile AlumnaSeal (RPI Corp., Mt. Prospect, IL) was pressed over the wells before the microplate was covered. The plate was placed on a rocking platform for 10 minutes to uniformly distribute the cells, and the microplate was placed in a tissue culture incubator for 72 hours. Each well was gently aspirated, fixed in 100% methanol for 10 minutes, and stained for 30 minutes with a 1% crystal violet solution in 50% methanol. The plate was gently washed, and 50 μL DMSO was added to each well for 30 minutes. The plate was analyzed in a Bio-Rad (Hercules, CA) microplate reader at 595 nm. The data were automatically recorded and transferred to an Excel spreadsheet in a microcomputer connected to the microplate reader.

The day before treatment, 4 × 10⁴ CHO cells were added to each microplate well in 200 μL F12 + 5% FBS and incubated. The next day, the cells were washed with HBSS and treated with the Na₂S·9H₂O in F12 without FBS in a total volume of 50 μL for 4 hours at 37°C, 5% CO₂. The wells were covered with sterile AlumnaSeal. After incubation, the cells were washed twice with HBSS and harvested with 50 μL of 0.005% trypsin + 53 μmol/L EDTA. The trypsin was inactivated with 70 μL F12 + FBS. A 10 μL aliquot was removed to measure acute cytotoxicity using 10 μL of 0.05% trypan blue vital dye in PBS. SCGE data were not used if the acute cytotoxicity exceeded 30%. The remaining cell suspension from each well was embedded in a layer of low-melting point agarose prepared with PBS on clear microscope slides that were previously coated with a layer of 1% normal melting point agarose prepared with deionized water and dried overnight. Detailed methods for preparing and electrophoresing the SCGE slides were published previously (31). Cellular membranes were removed by an overnight immersion in lysing solution at 4°C. The slides were placed in an alkaline buffer (pH 13.5) in an electrophoresis tank, and the DNA was denatured for 20 minutes. The slides were electrophoresed at 25 V, 300 mA (0.72 V/cm) for 40 minutes at 4°C. The slides were removed, neutralized with Tris buffer (pH 7.5), rinsed in cold water, dehydrated in cold methanol, dried at 50°C, and stored at room temperature in a covered slide box. For analysis, the slides were hydrated in cold water for 20 minutes and stained with 65 μL of 20 μg/mL ethidium bromide for 3 minutes. The slides were rinsed in cold water and analyzed with a Zeiss fluorescence microscope with an excitation filter of BP 546/10 nm and a barrier filter of 590 nm. For each experiment, two slides were prepared per treatment group. The slides were coded, and 25 randomly chosen nuclei were analyzed in each slide using a charge coupled device camera. A computerized image analysis system (Komet version 3.1, Kinetic Imaging Ltd., Liverpool, United Kingdom) was employed to determine the tail moment (integrated value of migrated DNA density multiplied by the migration distance) of the nuclei as an index of DNA damage. The digitalized data were automatically transferred to a computer-based spreadsheet for subsequent statistical analysis.

This method was done as developed by Wagner and Plewa.⁶

Briefly, the day before treatment, 4 × 10⁴ CHO cells were added to each microplate well in 200 μL F12 + 5% FBS and 100 μmol/L hydroxyurea and incubated. The next day, the cells were washed with HBSS and treated with the Na₂S in F12 with 100 μmol/L hydroxyurea and 100 μmol/L AraC without FBS in a total volume of 50 μL for 2 hours at 37°C, 5% CO₂. The wells were covered with sterile AlumnaSeal. After incubation, the protocol was the same for the SCGE assay described above.

Manipulations of toxic and mutagenic chemicals were conducted in certified biological/chemical safety hoods. For chronic cytotoxicity assays, the experiments were repeated at least twice with a minimum of eight independent replicates for each chemical concentration per experiment, and for the SCGE assay, three experiments were conducted with eight slides analyzed per treatment group. The median tail moment value for each slide was determined, and the data from all of the slides representing each Na₂S concentration were averaged. Averaged median values express a normal distribution according to the central limit theorem and were used with a one-way ANOVA test (62). A Holm-Sidak method for multiple comparisons versus the control group analysis was conducted.

We thank Tarannum Khan for valuable assistance and review of the article.