Tumors induce complex DNA damage in distant proliferative tissues in vivo

Abstract

That tumors cause changes in surrounding tissues is well documented, but whether they also affect distant tissues is uncertain. Such knowledge may be important in understanding the relationship between cancer and overall patient health. To address this question, we examined tissues distant to sites of implanted tumors for genomic damage using cohorts of C57BL/6 and BALB/c mice with early-stage subcutaneous syngeneic grafts, specifically, B16 melanoma, MO5076 sarcoma, and COLON26 carcinoma. Here we report that levels of two serious types of DNA damage, double-strand breaks (DSBs) measured by γ-H2AX focus formation and oxidatively induced non-DSB clustered DNA lesions (OCDLs), were elevated in tissues distant from the tumor site in tumor-bearing mice compared with their age- and sex-matched controls. Most affected were crypts in the gastrointestinal tract organs and skin, both highly proliferative tissues. Further investigation revealed that, compared with controls, tumor-bearing mice contained elevated amounts of activated macrophages in the distant gastrointestinal tissues, as well as elevated serum levels of several cytokines. One of these cytokines, CCL2/MCP-1, has been linked to several inflammation-related conditions and macrophage recruitment, and strikingly, CCL2-deficient mice lacked increased levels of DSBs and OCDLs in tissues distant from implanted tumors. Thus, this study is unique in being a direct demonstration that the presence of a tumor may induce a chronic inflammatory response in vivo, leading to increased systemic levels of DNA damage. Importantly, these findings suggest that tumors may have more profound effects on their hosts than heretofore expected.

The appearance of genome abnormalities and loss of viability in cells other than those directly hit with ionizing radiation (IR) is a well-documented process known as the radiation-induced bystander effect (1). Other studies reported that senescent and cancerous cells that have not been subjected to IR or other stresses also can affect normal cells sharing the same milieu (2, 3). These findings suggest that bystander effects (BE) are widespread phenomena in which cells undergoing any of a variety of stresses release substances that induce DNA damage in bystander cells. The bystander cells have been found to contain elevated levels of phosphorylated histone H2AX (γ-H2AX), indicating the presence of DNA damage, leading to increased and consequent genomic instability and decreased viability (4, 5).

Possible mediators of BE may be various inflammation-related cytokines that have been found at elevated levels in media conditioned on damaged, senescent, and cancerous cells. One, TGF-β, when added to cell cultures, induces elevated levels of DNA lesions similar to those induced by the conditioned media (3). Reactive oxygen species (ROS), including nitric oxide, have also been implicated in the transmission of BE (4, 6).

An important question is whether such effects demonstrated in vitro also exist in vivo. In classic radiobiology there is the so-called abscopal (out-of field or distant) effect, where irradiation of one organ results in a change in another, unirradiated organ (7). Although possibly caused by scatter from the main radiation source, abscopal effects may also suggest the presence of bystander-like processes in whole organisms.

Similar bystander-like effects have been reported in vivo for tissues neighboring injured, damaged, or stressed tissues. Elevated levels of inflammatory mediators, such as chemokines, cytokines, and prostaglandins (8), as well as elevated ROS levels (9–11), have been found in tumors and their immediate microenvironments. DNA damage has been reported in noncancerous cells neighboring tumors (12, 13), for example, in normal liver tissue adjacent to hepatocellular carcinoma (14).

In this study, we examined whether the presence of a tumor may cause systemic effects (i.e., affect tissues other than those directly proximal). We measured the levels of γ-H2AX, which generally marks double-strand breaks (DSBs) and eroded telomeres but also can be induced in the absence of DNA damage (15, 16), as well as oxidatively induced clustered DNA lesions (OCDLs) in tissues proximal and distant from nonmetastasizing implanted tumors in mice, and compared them to the levels in age- and sex-matched controls. These DNA lesions were found to be elevated in sites distant from the tumors, particularly in gastrointestinal tract (GIT) tissues, skin, and hair follicles. As expected, tumor-associated macrophages (17) were found in these mouse tumors, but in addition, increased amounts of activated macrophages were found in GIT tissues and skin of tumor-bearing mice compared with controls. To gain insight into possible mechanisms, we analyzed over 50 cytokines in the serum from tumor-bearing and control mice. One cytokine elevated in tumor-bearing mice, CCL2 [chemokine (C-C) ligand 2], was first described as a factor necessary for monocyte recruitment to sites of injury and macrophage differentiation, and has since been implicated in the pathogenesis of several inflammatory diseases (18–20). When implanted with tumors, CCL2-deficient mice failed to exhibit increased incidences of DNA damage, indicating that CCL2 is essential for this process. These findings are unique in demonstrating that the presence of an early stage nonmetastatic tumor may induce a level of systemic inflammation sufficient to induce potentially serious genomic damage in tissues distant from the tumor site.

Results

Three tumors were passaged through syngeneic donor mice, and tumor fragments were implanted s.c. into the experimental animals, B16 melanoma and M5076 reticulum sarcoma, each into 6 C57BL/6 (B6) mice, and COLON26 carcinoma into 6 BALB/c mice. All fragments formed subcutaneous masses in the test cohort (TST). Each TST cohort was compared with two sex- and age-matched control cohorts in independent experiments, a negative control of six PBS-injected mice (NC), and an acute inflammation control of six mice injected with 0.05 mL complete Freund's adjuvant (IC) (see schematic in Fig. S1). Approximately 2 wk postinjection, when the tumor masses approached 200 mg (≈7-mm length and 5-mm width), the animals were killed. Pathology analysis showed that the zones of the Freund's adjuvant injection were completely healed, with no signs of remaining skin inflammation.

The tissues were then analyzed by immunostaining for γ-H2AX foci and for OCDLs (comprised of at least one abasic or base lesion and a single-strand break in close vicinity) (21) in DNA. OCDLs were measured after digestion with enzymes that cleave the DNA at modified DNA residues, human APE1, which cleaves primarily at abasic sites, Escherichia coli EndoIII at oxypyrimidine sites, and human OGG1 at oxypurine sites (21). The number of sites was quantified by analysis of constant-field gel electrophoresis (22).

Affected Tissues.

In the pilot experiment, mice bearing B16 melanomas subjected to this protocol exhibited elevated levels of γ-H2AX foci in duodenum (2.30 ± 1.27 foci per cell; a 2.3-fold increase over controls) and in skin adjacent to the tumor (2.66 ± 0.04 foci per cell; a 3.5-fold increase over controls) (Tables S1 and S2). Other tissues examined including ovary, lung, liver, and kidney did not exhibit significant elevations in γ-H2AX levels.

To examine if the tumor-induced bystander DNA damage is caused by oxidative stress, we measured three types of OCDLs in the collected mouse tissues and found significant increases in the duodenum (0.33–0.36 clusters per mega-base pairs for all three types; 1.9-, 2.1-, and 2.8-fold increases over controls for abasic, oxypyrimidine, and oxypurine clusters, respectively) and skin surrounding the tumor (0.32 ± 0.04, 0.36 ± 0.06, and 0.54 ± 0.08 clusters per mega-base pair, 1.6-, 1.7-, and 1.6-fold increases over controls for abasic, oxypyrimidine, and oxypurine clusters, respectively) (Tables S1 and S3). Elevated levels of OCDLs appeared to be more widespread than DSBs in TST animals. Of the other tissues examined, ovary and lung exhibited significant increases in all three or two of the three types of OCDLs, respectively.

GIT and skin contain larger fractions of proliferating cells than the other examined tissues (23, 24), suggesting that S-phase cells may be more sensitive to DSB formation as reported in vitro for bystander DSB formation induced by IR (25, 26). Because DSBs may result if replication complexes encounter blocks resulting from non-DSB damage (15), greater numbers of OCDLs may have been converted to DSBs in duodenum and skin.

The oxypurine and oxyprymidine residues in OCDLs are signature oxidative lesions induced by an oxidizing cellular environment. Certain types of OCDLs may be particularly challenging to repair: for example, attempted simultaneous repair of two nearby lesions on opposite strands by base excision repair. Simultaneous DNA glycosylase cleavage on opposite strands can result in the production of a DSB, enhancing the biological impact of oxidatively induced DNA lesions (21, 27). In addition, the GIT, containing highly proliferative tissues, has been reported to be sensitive to oxidative stress (28, 29), a common source of non-DSB lesions.

Tumors Lead to the Induction of DNA Damage in GIT Organs and Skin.

To determine how widespread this phenomenon may be, we examined two other tumor models, M5076 reticulum cell sarcoma and COLON26 carcinoma. Because duodenum and skin of the B16 melanoma TST cohort exhibited elevated levels of both γ-H2AX foci and OCDLs, we focused these experiments on the analysis of GIT organs and regions of skin at various distances from the tumor. Compared with controls, TST cohorts exhibited elevated γ-H2AX focal frequencies in touch-print preparations of the four examined GIT tissues, duodenum, colon, rectum, and stomach, 1.4- to 2.3-fold for sarcoma-bearing mice and 1.7- to 3.7-fold for carcinoma-bearing mice (Fig. 1 A and E, and Tables S1 and S2). A pronounced accumulation of abasic sites and oxidized bases was found in all analyzed tissues of the GIT of the sarcoma- (1.8- to 4.6-fold) and carcinoma- (1.6- to 4.1-fold) bearing mice compared with control cohorts (Fig. 2, and Tables S1 and S3).

Fig. 1.

γ-H2AX foci are elevated in the GIT tissues and skin of COLON26 carcinoma-bearing mice. (A) (Left) GIT tissues consist of crypt structures (H&E-stained paraffin sections of NC mice), which contain continually renewing cells. (Right) γ-H2AX immunostaining in touch-print preparations of the noted organs from NC, acute IC, and TST cohorts. Images are representative maximum projections of optical sections with average numbers of γ-H2AX foci per cell. Higher-magnification images (Insets) more clearly show nuclei (red, counterstained with propidium iodide) containing γ-H2AX foci (green). (B) γ-H2AX staining in frozen duodenum sections. γ-H2AX levels are higher in the duodenum of TST mice compared with NC mice. (Insets) Crypt cells at higher magnification. (C) γ-H2AX staining in frozen sections of skin from control (NC and IC) mice compared with skin sections from tumor-bearing mice taken proximal (TST1) to the tumor mass (TST, tumor mass), about 0.5 cm (TST2) and 2 cm (TST3) from the tumor, with average numbers of γ-H2AX foci per cell. (D) Representative images of perpendicular hair follicle sections of NC and TST mice showing increased γ-H2AX staining in tumor-bearing mice. Graphs show the incidences of γ-H2AX foci in the GIT (E) and skin (F). The numbers 1, 2, and 3 correspond to the skin-sample location. White bars, NC; black bars, IC; gray bars, TST cohorts. Error bars are SDs (n = 6 mice per group). *Statistically significant difference in TST mice separately compared with both NC and IC controls (P < 0.01 for GIT tissues and P < 0.05 for skin). Magnification, 400× for main panels; 200× for H&E panel.

Fig. 2.

OCDLs in the GIT and skin from NC, IC, and TST cohorts of COLON26 carcinoma-bearing mice. In TST mice, OCDL levels were measured in the tumor (diagonally marked), as well as in normal skin proximal to tumor (checkered), close to tumor (solid), and far from tumor (horizontally marked). Clusters measured: (A) hAPE1 (abasic), (B) EndoIII (oxidized pyrimidine), and (C) hOGG1 (oxidized purine). Error bars are SD (n = 6) and three to five independent measurements. *Statistically significant difference in TST mice compared with both NC and IC controls (P < 0.01).

The GIT tissues consist of crypt structures (Fig. 1A, H&E panel), with rates of cell turnover among the highest in the body. Intestinal crypts contain proliferating stem cells and transit-amplifying cells in the basement layer, which then migrate to the walls of the crypts and villi, becoming differentiated specialized cell types (30). In touch-print preparations, γ-H2AX foci can be accurately counted in the cells (Fig. 1 A and E), but the origin of the cells within the tissue is unclear. In contrast, in frozen sections the origin of the cells in the tissue is clear, but γ-H2AX foci often cannot be accurately counted because a nucleus may be incomplete or may overlap one another (Fig. 1B). As an alternative to counting foci, the total intensity of γ-H2AX staining was measured in different areas of the colon and duodenum sections (duodenum images are presented in Fig. 1B). In the duodenum of carcinoma-bearing mice, the values of total γ-H2AX staining were threefold higher at the base of the crypts, where proliferating cells are located, than at the apexes; in the NC mice, the values were only 1.3-fold higher at the base compared with the top of the crypts. The ratio of this value, 2.3, is somewhat greater than the value, 1.7, obtained by focus-counting in touch-prints because the numbers of γ-H2AX foci in touch-prints are averaged over all of the cells, encompassing both cells with larger numbers of foci from the base of the crypts and cells with fewer foci from the top. Because we did not use specific markers to identify stem cells, we cannot conclude if the elevated γ-H2AX staining at the base of the crypts was stem cell-associated.

In addition to the GIT cells, mouse dermal fibroblasts were analyzed for γ-H2AX and OCDL levels. The low density of these cells in frozen sections permitted counting of γ-H2AX foci in skin samples proximal to the tumor and 0.5 cm distant from the tumor (in the sarcoma and carcinoma experiments), as well as 2 cm distant from the tumor (in the carcinoma experiment) (Fig. 1C). γ-H2AX levels in all three locations exhibited statistically significant increases in TST mice compared with controls (1.5- to 2.5-fold) (Fig. 1 C and F and Tables S1 and S2). OCDL levels also increased significantly (1.8- to 4-fold) in skin samples from all three locations (Fig. 2 and Tables S1 and S3).

Skin also contains hair follicles (Fig. S2 A and B), highly proliferative miniorgans whose cells are profoundly influenced by environmental conditions. Elevated levels of γ-H2AX staining were notable in both transverse (Fig. 1D) and longitudinal (Fig. S2C) follicle sections in all three examined skin locations of TST mice compared with controls. Measuring total intensity of γ-H2AX staining in various regions in longitudinal sections yielded values 1.5-fold elevated in the base bulb region (example presented in Fig. S2C), an area of high cell proliferation, in TST mice; those of the NC mice showed no difference.

Thus, this study is unique in demonstrating that the presence of a tumor can induce DNA damage in tissues of the organism distant from the tumor mass. In contrast to the two proliferative tissues, GIT and skin, no reproducible increases in γ-H2AX focal frequencies were observed in splenocytes or whole-blood lymphocytes (Table S2), both of which are composed primarily of resting cells. This finding also supports the notion that the BE is linked to DNA replication (25, 26).

Tumor-Induced DNA Damage Is CCL2-Mediated.

For damage to appear in tissues distant from the site of a tumor, molecules, possibly ROS, cytokines, or cells, must migrate from the tumor to these other tissues. We examined sections of colon, duodenum, and skin from the control and TST (M5076 sarcoma and COLON26 carcinoma) cohorts of mice for the presence of several types of immune cells, F4/80⁺ mature macrophages, CD3⁺ T lymphocytes, and CD45⁺ B-lymphocytes. Of these three cell types, only F4/80⁺ macrophages were found to be elevated in the tissues from TST animals (Fig. 3 A and C). F4/80⁺ cells are present in the colon as multifocal staining; in duodenum they are located in lamina propria; in skin they are often at the periphery of the neoplasm and in both epidermis and dermis (dermal layers are shown).

Fig. 3.

(A) Elevated numbers of F4/80⁺ macrophages in colon, duodenum, and skin of M5076 sarcoma-bearing mice (TST) compared with NC mice as determined by immunoperoxidase staining of paraffin sections. (B) In tumor mass and mucosa-associated lymphoid tissue (MALT), F4/80 expression is often severe, and tumors are infiltrated with CD3⁺ and CD45⁺ cells. Nuclei are counterstained with propidium iodide. Magnification, 200×. (C) Mature/activated macrophage staining intensity determined by immunoperoxidase staining for F4/80 in colon, duodenum, and skin of M5076 sarcoma-bearing mice (TST, red) compared with control cohort (NC, black). F4/80 intensity grades were 1, minimal; 2, mild; 3, moderate. Each dot in each group represents data from one animal. F4/80⁺ cells in lymphoid aggregates were excluded from the analysis. Horizontal bars represent medians. (D) CCL2 levels are higher in serum of WT B6 tumor-bearing mice over corresponding controls, but not in CCL2 KO mice. CCL2 levels are shown relative to the NC cohort (white). TST1, M5076 sarcoma-bearing mice (black); TST2, B16 melanoma-bearing mice (light gray). Error bars are SDs (n = 3–6). (E) The incidences of γ-H2AX foci increase in colon and duodenum of WT B6J tumor-bearing mice (TST1, TST2) compared with control mice (NC, IC), but not in CCL2 KO mice. Error bars are SDs (n = 5). *Statistically significant difference in a TST mice compared with both NC and IC controls (P < 0.05). (F) OCDLs are increased in colon and duodenum of B6J WT but not of CCL2 KO mice. Error bars are SDs (n = 5). *Statistically significant difference in TST mice compared with both NC and IC controls (P < 0.05).

We further measured the serum levels of 59 cytokines by ELISA in the B16 melanoma and COLON26 carcinoma-bearing mice. Of these, four factors were substantially (>3.5-fold) elevated above control values in both cohorts, CCL2 (MCP-1, monocyte chemoattractant protein-1), CCL4 (MIP-1β, macrophage inflammatory protein-1β), CCL7 (MCP-3), and CXCL10 (IP-10, inducible protein-10) (Fig. 3D and Fig. S3).

We focused our attention on CCL2, which is associated with rheumatoid arthritis, atherosclerosis, asthma, ulcerative colitis, and multiple sclerosis (31), and has been implicated in the progression and prognosis of several cancers (32). It has been reported to be secreted by tumor cells, normal tissues, and immune cells (33), acts as an attractant for monocytes, lymphocytes, NK cells, and immature dendritic cells to sites of injury, and is involved in macrophage activation (20). CCL2 is a major source of cyclooxygenase (COX)-2 (18), and is involved in TGF-β up-regulation (34), both main players in bystander signaling and carcinogenesis (6).

Thus, we repeated our standard protocol using CCL2-deficient mice implanted with WT M5076 reticulum cell sarcoma and B16 melanoma. Strikingly, no elevation of γ-H2AX foci or OCDLs was found in GIT tissues in these mice, demonstrating that this process is dependent on the presence of the CCL2 gene in the host (Fig. 3 E and F). When the CCL2 serum levels were assayed in the tumor-bearing KO mice, no measurable amounts were detected (Fig. 3D), suggesting that immune cells of the host rather than tumor cells are the main CCL2 producers.

Discussion

This report indicates that the presence of early stage nonmetastatic tumor may have more widespread and serious consequences on the host than previously documented. The findings demonstrate that tumors of different origin induce potentially dangerous oxidative DNA lesions in tissues distant from the tumor site, substantiating the existence of the tumor-induced BE in vivo. These findings also suggest that a source of persistent inflammation is required for accumulation of DNA damage, because mice subjected to acute inflammation (Freund's adjuvant-injected) as well as PBS-injected mice did not exhibit elevated DNA damage levels in distant tissues. Although the average elevation over the controls is approximately twofold for both OCDLs and γ-H2AX foci, some cells may harbor considerably more damage than the average values if the process creating the DNA damage is stochastic. Furthermore, the susceptibility of a cell population to genomic instability and oncogenic transformation may increase with the number of cells containing elevated DNA damage, resulting in long-term deleterious cellular and organismal outcomes.

Although the elevation of DNA damage levels was most obvious in the GIT and skin, significantly elevated levels of OCDLs were found in several other tissue types that lacked significantly elevated levels of γ-H2AX foci. The wider distribution of OCDLs relative to γ-H2AX foci may be caused by different fractions of proliferating cells in different tissues. DSB formation is possible in proliferating cells where replication forks interacting with isolated oxidative lesions may form a DSB, but OCDLs form independently of the cellular replicative state. GIT and skin both contain large proliferative fractions, perhaps making them more sensitive for DSB formation. These results suggest that the elevated DNA damage levels may be present in tissues throughout the organism, the levels perhaps depending on the robustness of cellular defense processes, such as the efficiency of DNA repair, the antioxidant capacity of the tissues, or their extent of infiltration by immune cells. Of the tissues examined in this study, two (namely liver and kidney) did not exhibit elevated OCDL levels in TST animals. We examined these two tissues and found that, in contrast to GIT and skin, kidney tissues in both control and TST animals were devoid of F4/80⁺ mature macrophages, a finding that strengthens the association between macrophages and OCDLs (Fig. S4A). On the other hand, liver tissues exhibited substantial numbers of F4/80⁺ macrophages, even in control animals, and further elevated numbers in tumor-bearing animals (Fig. S4A). Liver also exhibited manyfold greater antioxidant capacity than any of the other examined tissues (Fig. S4B), suggesting that the low levels of OCDLs in liver in the presence of F4/80⁺ macrophages may be a result of ample antioxidant defense.

There are several possible scenarios by which the presence of a tumor may lead to elevated levels of DNA damage in distant tissues. The tumors were found to be infiltrated with matured macrophages and lymphocytes (Fig. 3B), which could be activated by contact with the tumor cells to secrete a variety of factors. In addition, the tumor cells themselves are known to be capable of secreting a variety of cytokines and other factors (35), which may also activate resident or distant immune cells. However, the lack of any increase in oxidative lesions in CCL2 KO tumor-bearing mice suggests that, although tumor cells are capable of generating DNA damage-inducing signals, it is more likely that factors released by the immune cells are responsible for the distant DNA damage. Conversely, immune cells themselves, activated in the tumor mass, may move throughout the body and secrete substances in various tissues that induce DNA damage. Although some agents, such as ROS, secreted by the tumor may be too unstable, macromolecules and other signaling chemicals could be transported by the blood and either induce processes in distant tissues to generate DNA damage, or activate immune cells, perhaps already present in distant tissues for immune surveillance, which then secrete chemicals inducing DNA damage. Skin and GIT tissues in the tumor-bearing mice contained elevated levels of activated macrophages, suggesting that these immune cells may be involved in the local DNA damage induction in the tissues. Additionally, mucosa-associated lymphoid tissue is often highly amplified in chronically inflamed GIT (36), and may be another source of activated macrophages (Fig. 3B) and ROS.

The skin areas examined were either adjacent to the tumor or at some distance away. In the adjacent skin samples, transport of signals through gap-junctions could also contribute to the induction of DNA damage, as suggested by the gradual decrease of γ-H2AX foci in skin samples more distant from a tumor. Oxidative damage has been shown to increase in normal tissues bordering tumors (14), and cell-cell transmission of ROS and other metabolites through gap-junctions has been suggested as one possible signaling pathway for the radiation-induced bystander effect in vivo (37).

CCL2 up-regulation may be part of a general mechanism of tumor response (38) and blocking CCL2 may enhance immunotherapy (39). Our findings suggest that CCL2 may act with tumors to create a systemic inflammatory environment, which induces DNA damage in distant tissues throughout the organism. Because no obvious signs of acute inflammation were detected in skin and GIT of tumor-bearing mice, the involvement of CCL2 and macrophages in tumor-induced DNA damage suggests similarities with chronic tissue stress responses (40). In contrast to acute inflammatory responses, the chronic response may involve a shift in tissue homeostasis. Our findings link this type of chronic tissue response to the induction of oxidative DNA damage in normal tissues throughout the body, and provide a possible explanation for some of the consequences of chronic inflammation, including elevated cancer risk.

Materials and Methods

Additional details are provided in SI Materials and Methods.

Mice and Tumors.

Six-week-old C57BL/6NCr (B6) and BALB/cAnNCr (BALB/c) female mice were obtained from the Animal Production Area, National Cancer Institute (NCI)–Frederick. Six-week-old Jax MCP-1/CCL2 KO mice on the C57BL/6J background and WT female mice were purchased from the Jackson Laboratory. Cryopreserved murine B16 melanoma (host strain: B6), M5076 reticulum cell sarcoma (host strain: B6), and COLON26 carcinoma (host strain: BALB/c) were obtained from the Division of Cancer Treatment and Diagnosis tumor repository, NCI-Frederick, and were passaged through syngeneic “donor mice” before implanting into the experimental animals. Fragments of B16 melanoma and M5076 reticulum sarcoma were implanted s.c. into B6 mice, and COLON26 carcinoma was implanted into each BALB/c mice. The experimental scheme is presented in Fig. S1. In each of three experiments, three cohorts of mice were used: (i) TST: six animals were implanted with minced fragments of tumor harvested from donor mice; (ii) IC: six animals were subjected to a single s.c. injection of 0.05 mL complete Freund's adjuvant to determine whether any response observed was because of an acute inflammatory response rather than tumor growth; and (iii) NC: six animals were injected s.c. with sterile PBS. Mice were killed with CO₂ when grafted tumors in the TST cohort reached a volume of ≈200 mg (early-stage nonmetastatic malignancies), and tissues were harvested for analysis.

DNA Damage Analysis.

γ-H2AX immunostaining in tissue touch-print preparations and frozen sections, as well as in cytospin samples of lymphocytes and splenocytes, was performed as previously described (41). The primary rabbit polyclonal anti-γ-H2AX antibody (Novus Biologicals) and the secondary goat anti-rabbit Alexa-488-conjugated IgG (Invitrogen) were used. Nuclei were stained with propidium iodide. Laser scanning confocal microscopy was performed with a Nikon PCM 2000 (Nikon, Inc.).

For OCDL detection, DNA isolated from the tissues was digested with human repair enzymes APE1 and OGG1, and E. coli EndoIII (New England Biolabs). For each enzyme-treated sample, a corresponding nonenzyme containing sample was also run as a control following the same steps. An adaptation of constant-field gel electrophoresis was used (42). Electronic images for each gel lane were processed using QuantiScan (BioSoft). DNA standards (λ-HindIII Digest) were used to obtain the corresponding dispersion curve with Origin 6.1 (OriginLab). The numbers of average lengths (L_n) for each sample were calculated using the equations described in ref. 42. The frequencies of OCDLs were measured based on the L_n values of the enzyme-treated sample (+ lane) and the accompanying control sample (− lane).