Ulcerative Colitis–Associated Colorectal Cancer Arises in a Field of Short Telomeres, Senescence, and Inflammation

Ulcerative colitis (UC) is an inflammatory bowel disease that predisposes to the development of colorectal cancer. The sequence of pathologic changes observed in the colonic epithelium progresses from nondysplastic (no pathologic abnormalities) to dysplasia [indefinite dysplasia, low-grade dysplasia (LGD), and high-grade dysplasia (HGD)] and to cancer. These neoplastic changes can be widespread and/or multifocal. Molecular abnormalities, including telomere shortening, chromosomal instability, p53 alterations, and aneuploidy, underlie these neoplastic changes in a multistep process of cancer progression (1). Remarkably, these molecular alterations are found not only in neoplasia but also in histologically normal tissue, indicating that they precede the development of dysplasia (2). The presence of this “field effect” is common in epithelial carcinogenesis (3). In UC, we and others have reported a field effect for aneuploidy, p53 mutations, and clonal chromosomal abnormalities in normal epithelium adjacent to dysplasia (4–7) or even throughout the whole colon (8). However, the cellular and molecular mechanisms that underlie this field effect are poorly understood.

Telomeres shorten as a consequence of cell replication and oxidative damage (9). We have previously demonstrated that there is an accelerated shortening in UC colonocytes compared with normal individuals (10) and that telomere shortening is one of the initial triggers of UC tumorigenesis, which leads to chromosomal instability through cycles of bridge–breakage–fusion (11). Thus, we hypothesized that UC is a disease of accelerated colon aging, in which chronic inflammation leads to telomere shortening and eventually, to cancer progression. Evidence for an association between chronic inflammation and cancer risk comes from the fact that the duration, extent, and severity of inflammation are risk factors for neoplasia in UC (12) and that anti-inflammatory medications reduce that risk (13). Several studies suggest that reactive oxygen species produced by activated neutrophils and macrophages can contribute to tumor initiation by causing DNA damage (reviewed in refs. 14, 15). However, the mechanistic link between telomere shortening, inflammation, and neoplastic progression has not been fully explored.

In the presence of proficient cellular checkpoints, the outcome of telomere shortening is not chromosomal instability but senescence, a state of irreversible growth arrest that acts as a tumor suppressor mechanism by preventing the replication of damaged cells (16). Short, dysfunctional telomeres are sensed by the cell as DNA double-strand breaks and thus trigger the DNA damage response through phosphorylation of the histone H2AX (γ-H2AX; 17). Consistent with its role as a tumor suppressor mechanism, in vivo senescence has been detected in premalignant lesions but lost in cancers (18). This suggests that senescence markers, such as Dec1, a p53-regulated senescence effector (19), could be useful in assessing cancer risk. Interestingly, senescence also has a tumorigenic effect, as senescent cells secrete a variety of cytokines and other proinflammatory proteins that promote tumor progression (20). This intriguing link between senescence and inflammation is currently the subject of active research (21), but little is known of its role in in vivo cancer progression.

We postulated that in UC (1), telomere shortening and DNA damage lead to cellular senescence in preneoplastic fields (2); at some point in the dysplastic sequence, senescence is bypassed to allow tumor progression (3); and chronic inflammation is the underlying mechanism that triggers telomere shortening and senescence in the preneoplastic colon of UC patients. We addressed these issues by analyzing telomere length, telomerase, DNA damage, senescence, p53, p16, and inflammation in multiple biopsies from all histologic grades collected along the colon of UC patients.

This study included multiple biopsies collected from surgically resected colons from 9 UC patients with HGD or cancer (progressors) and 4 UC patients without dysplasia or cancer (nonprogressors; Table 1). The indication for colectomy in UC progressors was diagnosis of HGD or cancer at colonoscopy, whereas for UC nonprogressors it was intractability of symptoms. All patients had at least 8 years of disease duration. For each progressor, an average of 7 biopsies was analyzed (minimum 5, maximum 10). These biopsies included HGD, LGD, and nondysplastic biopsies collected at random locations in the colon from each patient. For each biopsy, the position in centimeters from the rectum was recorded. Colon maps of the histologic diagnoses of progressors are included in Supplementary Figure S1. For each nonprogressor, 5 biopsies negative for dysplasia were collected from locations throughout the colon. As controls we included 21 colon biopsies collected at colonoscopy from individuals without UC (non-UC colon). The diagnoses at colonoscopy of the normal controls included diverticulitis, prolapse, constipation, hyperplastic polyps, liposarcoma, impacted fecalith, and normal colon cancer screening. All biopsies were split in thirds: the first third was used for epithelial isolation (see below), the second third was lightly fixed in 4% paraformaldehyde and paraffin embedded for immunofluorescence studies, and the last third was frozen for future use. Formalin-fixed, paraffin-embedded biopsies that were routinely collected at colectomy and matched the locations of the frozen biopsies were used for immunohistochemical (IHC) staining. Samples were collected at the University of Washington Medical Center, Seattle, WA, and at the Cleveland Clinic Foundation, Cleveland, OH. These studies were approved by the Human Subjects Review Boards of each institution with annual renewals.

Epithelial cells from frozen biopsies were isolated using epithelial shake-off as previously described (8). Cytokeratin staining confirmed that at least 90% of the cells were epithelial. Epithelial and stromal cells were treated with 1× CHAPS lysis buffer (Chemicon) for 30 minutes on ice. After centrifugation, the supernatant, which included the protein extract, was immediately frozen. The pellet was used for DNA extraction with Qiagen DNA extraction kits (Qiagen), according to the manufacturer's instructions.

Telomere length was measured by quantitative PCR (Q-PCR), as previously described (22). Telomeric DNA and a single-copy internal control gene (36B4, acidic ribosomal phosphoprotein PO) were amplified in each sample. The amount of telomeric DNA (T) was divided by the amount of single-copy internal control gene DNA (S), producing a relative measurement of the telomere length (T/S ratio). Two control DNA samples were included in each run to allow for normalization between experiments, and periodic reproducibility experiments were performed to guarantee accurate measurements. The intra- and interassay variability (coefficient of variation) for Q-PCR was 6% and 7%, respectively.

Telomerase activity was measured by Q-PCR, using a modified protocol of the real-time TRAP assay, as previously described (23). See Supplementary Material for detailed protocol.

Paraffin-embedded, lightly paraformaldehyde-fixed slides were processed using a modification of previous protocols (11, 24). Briefly, after antigen retrieval, slides were stained with a mouse monoclonal anti-phospho-histone H2AX antibody (Ser139) clone JBW301 (Upstate Biotech) and rabbit anti-Dec1 (a generous gift from Dr. Adrian Harris). Slides were incubated with secondary antibodies and counterstained (see Supplementary Material for detailed protocol). Images were taken on a Zeiss LSM Meta 510 microscope at 63×, with excitation at 543 nm for Alexa 568, 633 nm for Alexa 647, and 405 nm for DAPI, using sequential scans at constant settings for all slides. Images were quantitatively analyzed by calculating the average γ-H2AX and Dec1 staining intensity of at least 100 nuclei per biopsy in a minimum of 3 fields (25, 26).

Paraffin-embedded, formaldehyde-fixed slides were stained with a mouse monoclonal anti-p53 antibody clone DO-7 (Dako) or a rabbit polyclonal anti-p16 antibody (Santa Cruz Biotechnology) after antigen retrieval. Slides were washed, incubated with secondary antibodies, fixed, counterstained, and mounted. Images were acquired with an Olympus BX41 microscope and processed with a Nuance Multispectral Imaging System (CRI). This technique allows the deconvolution of the intensities of multiple chromogens by acquiring a spectral absorbance curve at every pixel of the image. The mean density of each chromogen was quantified after segmentation of nuclear or cytoplasmic areas, using a watershed algorithm with software developed in our laboratory (26).

Formalin-fixed, paraffin-embedded slides were stained with hematoxylin–eosin (H&E) and examined under a light microscope. Chronic and acute inflammation was analyzed. Chronic inflammation was assessed by 2 different parameters: (i) the presence of infiltrating leukocytes in lamina propria and (ii) the presence of lymphoid aggregates in mucosa and submucosa. Infiltrating leukocytes were quantified using a semiquantitative scale from 1 to 3 and lymphoid aggregates, using a binary scale with 0 for absence and 1 for presence. Acute inflammation was measured using the conventional pathologic score for inflammation activity: 0, inactive; 1, UC cryptitis; 2, UC, crypt abcesses; 3, UC, numerous crypt abcesses; and 4, UC ulcerated and granulation tissue. In a subset of slides, T cells, B cells, and macrophages were stained with anti-CD3, anti-CD20, and anti-CD68 antibodies, respectively, using standard immunohistochemical techniques. They were quantified using a semiquantitative scale from 1 to 3. Plasma cells were quantified in the same manner from H&E slides.

Nonparametric tests were used throughout the studies because of the small number of cases in some groups. In particular, the Mann–Whitney test and the Kruskal–Wallis test were used wherever appropriate to compare the means of quantitative variables in groups defined by qualitative variables. The strength of the associations among quantitative variables was presented by Spearman's correlation coefficient. Chi-squared tests were used to compare qualitative variables. All the tests were 2-sided at an α level of 0.05.

The analysis of multiple biopsies, mostly negative for dysplasia, along the colon of progressors allowed us to accurately map the molecular and cellular abnormalities studied in relation to their proximity to dysplasia. Detailed colon maps for each of the 9 progressors are shown in Supplementary Figure S1. The information derived from biopsies analyzed for non-UC colon, UC nonprogressors, and UC progressors is summarized in a heat map in Supplementary Figure S2, with samples ordered by increasing histologic grade (negative, LGD, and HGD) and by decreasing distance (in centimeters) to the nearest site of HGD or cancer. One of the most striking observations from this heat map is that shorter telomeres and higher levels of γ-H2AX, Dec1, and infiltrating leukocytes appeared more prevalent in negative biopsies located less than 10 cm from HGD or cancer than in biopsies located further away. For subsequent analyses, we grouped biopsies according to these categories. Figure 1 shows the comparison of means for all the quantitative parameters of the study using this biopsy grouping.

Early UC progression was significantly associated with telomere shortening: a dramatic telomere length decline was observed in colonic epithelium from progressors compared with nonprogressors and for progressors between nondysplastic biopsies away from dysplasia and closer to dysplasia. The shortest telomeres in progressors were found in LGD biopsies (Fig. 1 and Supplementary Table 1). Overall, nondysplastic biopsies of progressors had, on average, telomeres that were 22.1% shorter than the telomeres of nonprogressors (mean ± SD: 1.02 ± 0.18 for nonprogressors vs. 0.80 ± 0.19 for progressors; Mann–Whitney P < 0.0001, Supplementary Table 1). Interestingly, stromal telomere length decreased significantly only between biopsies from UC patients compared with non-UC colon (Fig. 1B and Supplementary Table 1). Throughout the neoplastic progression of UC, the telomeres of stroma cells remained unchanged and significantly longer than the telomeres of epithelial cells (mean telomere length ± SD for all UC biopsies measured in stroma 1.10 ± 0.25 vs. 0.84 ± 0.22 measured in epithelium; Mann–Whitney P < 0.0001).

A significant decrease of telomerase activity was found between nonprogressor epithelium and progressor nondysplastic epithelium that was more than 10 cm away from neoplasia (Fig. 1C and Supplementary Table 1). Telomerase activity in the stroma was not statistically different between groups (Fig. 1D and Supplementary Table 1).

In addition to shorter telomeres, negative biopsies less than 10 cm to HGD or cancer also showed significantly higher levels of DNA damage, as quantified by γ-H2AX, and senescence, as quantified by Dec1, compared with biopsies located at 10 cm or more from those dysplastic areas (Fig. 1E and F and Supplementary Table 1).

Negative biopsies from UC progressors showed significantly lower expression levels for p16 and p53 than negative biopsies from UC nonprogressors (Supplementary Table 1). However, there was no difference in the expression levels of either marker between biopsies located less than or more than 10 cm away from dysplasia. It is important to note that the high expression of p16 and p53 in UC nonprogressors was not statistically significantly different from the levels of the normal controls since some of the normal colons also showed high levels of both markers. The different underlying medical indications for colectomies in those “normal” colons may be the reason for those unexpected positive results.

While early UC progression was associated with telomere shortening, the pattern reversed in late UC progression, as telomeres lengthened in the transition from LGD to HGD. Telomeres were the shortest in LGD (0.62 ± 0.19) but significantly increased in HGD (0.83 ± 0.21, P = 0.039) to a length comparable with nondysplastic epithelium that is away from neoplasia (Fig. 1B and Supplementary Table 1). The most likely explanation for this lengthening is reactivation of telomerase. We did observe an increase of epithelial telomerase activity in the transition from LGD to HGD (Fig. 1D), but this difference did not reach statistical significance (Supplementary Table 1).

In addition to telomere lengthening, the later stages of UC neoplastic progression also involved a decrease of senescence, as indicated by the levels of Dec1. Expression of Dec1 was the highest in LGD biopsies (1.30 ± 0.25), but in HGD biopsies decreased to the level of nondysplastic biopsies away from neoplasia (1.04 ± 0.10, P = 0.045; Fig. 1G and Supplementary Table 1). The expression levels of p16 and p53 progressively increased with progression (Fig 1H and 1I), although the high heterogeneity observed in both inter- and intrabiopsy specimens rendered the comparison of means nonsignificant (Supplementary Table 1).

p16 and p53 expression levels were strongly associated when all UC biopsies (from UC nonprogressors and UC progressors) were included in the analysis (Table 2). This association remained significant at the different stages of progression, that is, UC nonprogressors, UC progressors negative for dysplasia biopsies, and UC dysplastic biopsies (LGD + HGD). In addition, when all UC biopsies were considered, high levels of γ-H2AX were strongly associated with Dec1 and with shorter telomeres. Interestingly, in UC nonprogressors, high levels of γ-H2AX were also associated with increased expression of p16 and p53. However, in nondysplastic biopsies from UC progressors, high DNA damage was associated only with increased senescence but not with p16 or p53. Finally, in UC dysplastic biopsies, longer telomeres were correlated with decreased Dec1 and, conversely, with increased p53 expression.

On the basis of previous data, we hypothesized that the frequency of alterations in histologically negative biopsies should be associated with the extent of dysplasia throughout the colon. To test this hypothesis, we quantified the percentage of colonic dysplasia, including LGD, HGD, and cancer, for each of the 9 progressors. This percentage ranged from 5.2% to 61.3% of the biopsies collected at colectomy [an average of 109 (minimum 86, maximum 134) biopsies per patient]. Then we examined the correlation of percentage of dysplasia with the mean value for telomere length, telomerase activity, and γ-H2AX, Dec1, p16, and p53 expression calculated from all the histologically normal biopsies present in the colon. Overall, patients with greater extents of dysplasia had shorter telomeres (P = 0.049, Fig. 2A), more DNA damage (P = 0.002, Fig. 2C), and a tendency for more senescence (P = 0.076, Fig. 2D) and more p53 expression (P = 0.087, Fig. 2F) elsewhere in the colon. This is likely due to the fact that the preneoplastic field encompasses a larger region in these patients, predisposing them to a higher chance of progression to dysplasia.

Figure 3A–F illustrates the scoring system used for infiltrating leukocytes in the lamina propria. We observed that LGD biopsies and negative biopsies less than 10 cm from dysplasia frequently showed very high levels of infiltrating leukocytes (3+; Fig. 3D–E), which were rarely observed in other biopsies (Fig. 3A–C and F). When we quantified the proportion of biopsies with high levels of infiltrating leukocytes (3+), the association with progression was highly significant (P = 0.017, Table 3). Around half of the biopsies in the 10-cm preneoplastic field and 62.5% of LGD biopsies showed abundant infiltrating leukocytes, while in the other groups only 20% or less of the biopsies showed this feature. Remarkably, none of the HGD biopsies had 3+ infiltrating leukocytes. The 2 additional inflammation parameters assessed in this study, lymphoid aggregates and inflammatory activity, did not show any association with UC progression (Table 3).

The striking association between lamina propria–infiltrating leukocytes and UC tumorigenesis prompted us to analyze these biopsies for specific subsets of leukocytes. The leukocytes expanding the lamina propria appeared to be a mixture of T cells, B cells, macrophages, and plasma cells, and this composition did not differ between biopsies with low or high lamina propria infiltration (Fig. 3G–I: nonprogressor, low infiltration; Fig. 3J–L: LGD, high infiltration). However, there appeared to be a difference in the location of B cells and T cells, which were basal in nonprogressor biopsies (Fig. 3G–H) but spread out in the mucosa of progressor biopsies (Fig. 3J–K).

Next, we explored to what extent the molecular alterations found in the preneoplastic field were related to the inflammatory process that underlies this disease. The levels of the 6 molecular parameters included in this study were compared between UC biopsies with low or high inflammation as defined by the presence of lamina propria–infiltrating leukocytes, lymphoid aggregates, and the acute inflammation index. Interestingly, the infiltrating leukocytes showed several highly significant associations with molecular parameters (Fig. 4). UC biopsies with high levels of infiltrating leukocytes in the lamina propria had significantly shorter telomeres in both epithelium and stroma (P < 0.0001 and P = 0.0001, respectively; Fig. 4A and B). Moreover, they had increased stromal telomerase activity (P = 0.027; Fig. 4D), more senescence as measured by Dec1 levels (P = 0.006; Fig. 4F), decreased p53 expression (P = 0.032; Fig. 4H), and a tendency for higher DNA damage (P = 0.06; Fig. 4E).

The acute inflammation index and the lymphoid aggregate index showed only significant associations with telomerase activity; for the acute inflammation index the association was in the stroma (P = 0.037) and for the aggregate index the association was in the epithelium (P = 0.013). None of the other molecular alterations were different for biopsies with inflammation as measured by these 2 parameters (data not shown).

Our results show that in UC patients, dysplasia arises in a field of inflammation, short telomeres, DNA damage, and senescence and provides novel in vivo evidence for the role of senescence as a tumor suppressor mechanism. Although telomeres are shorter in nondysplastic biopsies of UC progressors compared with nonprogressors, the shortening is more pronounced in the 10 cm surrounding HGD or cancer. Within this area, DNA damage (measured by γ-H2AX) and senescence (measured by Dec1) are significantly increased, likely as a consequence of the underlying chronic inflammation, as demonstrated by the abundance of infiltrating leukocytes. Interestingly, LGD biopsies harbor the shortest telomeres and the highest levels of senescence and infiltrating leukocytes, while HGD biopsies display long telomeres, low levels of senescence, and reduced amount of infiltrating leukocytes. This suggests that telomere-induced senescence acts as a tumor suppressor mechanism in LGD and that this checkpoint is bypassed in HGD, allowing cell proliferation and subsequent cancer. The escape from senescence coincides with the reactivation of telomerase and the lengthening of telomeres, which could be one of the mechanisms that allow the release of the blockade. In addition, HGD biopsies show overexpression of p53 (this study and ref. 27), which is likely due to the stabilization of mutant p53 protein (28). Short telomeres induce senescence through the activation of p53; thus, the loss of this tumor suppressor allows a permissive environment in which critically short telomeres can generate chromosomal instability and promote tumor progression (29). The loss of functional p53 may also be responsible for the decreased expression of Dec1, as Dec1 is a transcription factor downstream from p53 (19). Thus, the decreased levels of Dec1 observed in HGD are concordant with the overexpression of p53.

In addition to p53, the induction of senescence by critically short telomeres also involves the activation of the p16 pathway (30). In this study, however, we did not observe an increase of p53 or p16 in parallel to the shortening of telomeres and the increase of Dec1 in histologically normal biopsies from UC progressors. On the contrary, the levels of both tumor suppressors were higher in UC nonprogressors than in histologically normal biopsies from progressors. A possible interpretation is that the pathways of activation of the DNA damage response are functional in UC nonprogressors but are somehow impaired in UC progressors. Previous findings of methylation of the p16 promoter (31) and p53 mutations (4) in nondysplastic biopsies of UC progressors support this hypothesis. Moreover, in UC nonprogressors γ-H2AX was associated with Dec1, p16, and p53, suggesting the activation of the DNA damage response and subsequent senescence in biopsies with DNA double-strand breaks. However, in nondysplastic biopsies from UC progressors, there was a strong association between γ-H2AX and Dec1 (both markers are higher closer to dysplasia) but no association of γ-H2AX and Dec1 with p53 and p16. These surprising results might be an indication of additional senescence pathways activated in the presence of a defective DNA damage response in histologically normal biopsies from UC progressors. The fact that overexpression of Dec1 induces senescence in p53-knockdown cells (19) supports this possibility, but further research is needed to clarify these results.

Another interesting aspect of this study is the association between infiltrating leukocytes in the lamina propria and progression to dysplasia, supporting a connection between inflammation and cancer. We measured 2 types of inflammation: active and chronic. Active inflammation is the classical pathologic index, which is based on the presence of granulocytes in the epithelium. Chronic inflammation refers to the infiltration of leukocytes (mainly lymphocytes) in the lamina propria. Interestingly, we found that chronic inflammation, but not active inflammation, is associated with tumor progression in UC. Moreover, abundant infiltrating leukocytes were associated with shorter telomeres, increased levels of γ-H2AX, increased levels of Dec1, and reduced p53 expression. These results support previous knowledge that tumor-infiltrating leukocytes promote tumor growth (32, 33) and suggest that this could be due to DNA damage, telomere shortening, and senescence induced by oxidative stress. Indeed, inflammatory cells produce large amounts of reactive oxygen species, which play a fundamental role in UC tumorigenesis (15, 34, 35).

In spite of accumulating evidence that chronic inflammation contributes to tumor progression in UC, the reason is still unknown; while all UC patients have chronic inflammation, only a subset will develop cancer. Individual variation in telomere attrition, DNA damage response, or oxygen-free radical scavenging might contribute to this risk. In this regard, this study points to 2 different, nonexclusive hypotheses. The first is that UC nonprogressors may be more efficient in mounting an effective DNA damage response than UC progressors, as previously discussed. The second hypothesis is that UC nonprogressors may be less prone to tumor progression thanks to higher telomerase activity, which has been reported to protect the cells from oxidative stress (36, 37). Both hypotheses merit further investigation from a mechanistic point of view as well as for the potential for p53 and p16 expression and telomerase activity to be useful markers for cancer risk in UC.

While the prevalent view is that chronic inflammation is the initial trigger for tumorigenesis in UC, there is the possibility that inflammation is also a consequence of underlying senescence, as senescent cells produce a senescence-associated secretory phenotype (SASP) that is proinflammatory. Among many other functions of SASP is the recruitment of leukocytes to the tissue (20), which suggests that the abundant infiltrating leukocytes observed in LGD could be a consequence of a high level of senescence in these biopsies. If that were the case, then high infiltration should diminish as senescence is reduced, as observed in HGD. While senescence acts as a tumor suppressor mechanism, the SASP is tumorigenic, as it modifies the tissue microenvironment to promote tumor progression. Interestingly, we observed that the stroma of biopsies with high infiltration showed significantly shorter telomeres and more telomerase activity than the stroma of biopsies with low infiltration. Moreover, although the proportion of leukocyte subsets did not seem to change with infiltration, their distribution might be affected, as indicated by the observed migration of B cells and T cells from the basal layer to the apical surface. Other studies have reported an important role for stromal alterations in UC carcinogenesis (38, 39). Nevertheless, the observations reported here are limited because of the small number of samples and need further investigation.

In conclusion, our results support a model in which inflammation, telomere shortening, and high levels of DNA damage activate the induction of senescence in the colon of UC progressors. Senescence acts as a tumor suppressor mechanism, preventing colonocytes from progressing further than LGD. Eventually, some colonocytes bypass senescence, coinciding with the lengthening of telomeres through activation of telomerase and the loss of p53 function. This allows colonocytes with genetic damage to proliferate and progress to HGD and cancer. Further investigation is needed to elucidate the role of p16 and p53 in the induction and further escape of senescence in UC cancer progression.

No potential conflicts of interest were disclosed.

We thank Faith Tierney, Jeanne Fredrickson, Jasmine Gallaher, Julie Mejia, Calvin Ngo, Trang Le, and Arielle Samuelson for technical support; Wen-Tang Shen for computer support; and Allyn Stevens, Yasuko Tamura, Jeanne Stanton, and Mallory Smith for research biopsy coordination.

This study is funded by NIH P20 CA103728, R01 CA068124, P30 AG13280, NIH DK56924, K07 CA137136, and the Crohn's and Colitis Foundation of America.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.