Complex engagement of DNA-damage response pathways in

Carcinogenesis vol.28 no.10 pp.2082–2088, 2007doi:10.1093/carcin/bgm108Advance Access publication May 22, 2007

Complex engagement of DNA damage response pathways inhuman cancer and in lung tumor progression

Paolo Giovanni Nuciforoy, Chiara Luisey, Maria Capra,Giuseppe Pelosi1 and Fabrizio d’Adda di Fagagna�

FIRC Institute of Molecular Oncology Foundation, via Adamello 16, 20139Milan, Italy and 1Department of Pathology, European Institute of Oncology,20139 Milan, Italy

�To whom correspondence should be addressed. Tel. þ39 02 574303.227;Fax þ39 02 574303.231Email: [email protected]

Tumor initiation and progression provide a multitude of occasionsfor the generation of DNA damage and the consequent activationof the DNA damage response (DDR) pathway. DDR signalinginvolves the engagement of key factors such as ATM, CHK2,53BP1 and the phosphorylation of histone H2AX (g-H2AX).The systematic study of DDR in human tumors and normal tissuesby high-throughput tissue microarrays revealed that ATM andg-H2AX were engaged in cancer but the extent of their activationwas strongly affected by the organ and cell type involved, whereas53BP1 loss was the most consistent feature among the tumorstudied. Unexpectedly, we also observed activated DDR markersin morphologically normal tissues, also in association with inflam-mation. Analysis of the dynamic engagement of DDR along thedifferent stages of lung tumorigenesis showed that 53BP1 lossoccurs early at the transition from normal to dysplastic changewhereas the activated forms of ATM and CHK2, but not g-H2AX,initially accumulate in pre-invasive lesions and are then lost dur-ing tumor progression. In individual lung tumors, the activationof ATM, CHK2 and the presence of 53BP1 were consistently cor-related, whereas g-H2AX did not correlate with activated ATM.Finally, the study of associations between critical clinicopatholog-ical parameters and activated DDR factors highlighted a statisti-cally meaningful correlation between reduced local tumorextension and the phosphorylation of ATM, CHK2 and the pres-ence of 53BP1, whereas no significant correlations with parame-ters such as survival or relapse of early-stage lung carcinomaswere found.

Introduction

Cells respond promptly to the generation of DNA damage by launch-ing a coordinated set of actions, collectively known as the DNAdamage response (DDR) (1). This program has the dual function ofactivating the mechanisms devoted to DNA damage repair and ofhalting cell cycle progression. The latter checkpoint function canact either transiently, providing the time to repair the damage, orpermanently, leading to a prolonged cell cycle arrest called cellularsenescence or to a cell suicide program termed apoptosis. The DDR isinitiated by the recognition of DNA damage by a set of sensors thatactivate the protein kinase activity of ATM, when the trigger is a DNAdouble-strand break (DSB), or ATR, when single-strand DNA hasbeen exposed (2). ATM and ATR can modify chromatin at the siteof DNA damage, by phosphorylating S139 of histone H2AX, an eventthat facilitates DDR enforcement. Activated ATM and ATR signal totwo more downstream kinases, CHK1 and CHK2, which modify ad-ditional elements of the DDR cascade. Signaling between upstreamand downstream kinases is facilitated by a set of mediators of the

DDR which include 53BP1, whose accumulation at the site of damageis mediated by chromatin modification events (3,4). DDR pathwaysare not strictly linear and redundant signaling occurs, making theDDR network more robust.

Tumors are the result of clonal amplification of cells carrying ge-netic alterations that drive their unrestrained proliferation. Tumori-genesis provides various occasions for DDR activation. These includetelomere shortening (5,6), reactive radical species accumulation (7),chromatin alterations (8), hypoxia (9) and rampant genome instability.In addition, oncogene activation per se is sufficient to engage theDDR machinery (10,11) and markers of an activated DDR have beendetected in early tumors (12,13).

Genetic evidence for an important tumor suppressive role of DDRgenes in vivo has been provided by the observation of a tumor-pronecondition of patients carrying mutations in various DDR genes such asATM, NBS1, ATR, MRE11, CHK2, BRCA1 and 2 and in knockoutmice for several DDR genes (14,15). In addition, mutations and lossof DDR gene expression have been reported in tumors (16–23). Re-cently, we and others have shown that a mechanism through whichDDR genes can exert their tumor suppressive functions is by enforc-ing cellular senescence, thereby preventing the proliferation of dam-aged cells (10,11).

It is still unclear how widespread DDR activation is among humantumors of different histotypes and at different stages during cancerprogression. Here, we report on the engagement of key components ofthe DDR-signaling cascade in a broad collection of human tumors ofdifferent histotypes. In addition, we show in detail the differentialdegree of activation of individual DDR factors during lung tumorprogression. In these same samples, we also analyze the correlationsamong activated DDR factors and the potential association betweenthe activation of individual DDR factors and various clinicopathologicparameters.

Materials and methods

Samples and tissue microarray preparation

Normal skin for antibody validation (Figure 1) was from a female patient whounderwent reductive mammoplasty. After surgical excision, the sample was cutinto two and one part only was ex vivo X-rays irradiated (20 Gy). After 30 minincubation in phosphate-buffered saline, both parts were fixed in bufferedformalin and included in paraffin. Normal lung samples (Figure 2C) werederived from three patients who underwent a surgical procedure for a non-neoplastic pathology. Tissue samples were provided by Istituto Europeo diOncologia (Milano).

Three multitumor tissue microarrays (TMAs) were specifically designed forthe screening (Table I), and prepared as previously described in ref. 24 withminor modifications. Briefly, two representative normal (when available) andtumor areas (diameter 0.6 mm) from each sample, previously identified onhematoxylin–eosin-stained sections, were removed from the donor blocks anddeposited on the recipient block using a custom-built precision instrument (Tis-sue Arrayer—Beecher Instruments, Sun Prairie, WI). Two micron sections of theresulting recipient block were cut, mounted on glass slides and processed forimmunohistochemistry (IHC). Overall, multitumor TMAs contained tissue sam-ples from breast (11 normal, 27 fibroadenomas and 43 carcinomas), lung (64normal and 107 carcinomas), colon–rectum (13 normal and 18 carcinomas),kidney (15 normal and 18 carcinomas), larynx (20 normal and 28 carcinomas),stomach (24 normal and 28 carcinomas), hematopoietic system (20 non-Hodg-kin’s and 8 Hodgkin’s lymphomas), skin (14 nevi and 20 melanoma), soft tissues(18) and bone (10 osteogenic sarcomas) and central nervous system (4 menin-giomas and 20 gliomas). Importantly, care has been taken to avoid proximity toa resected margin as we have observed that thermal damage during surgery mayinduce a strong DDR (Nuciforo PG, unpublished observation).

Specimens were provided by the Pathology Departments of Ospedale Mag-giore (Novara), Presidio Ospedaliero (Vimercate) and Ospedale Sacco (Milano).

Lung-specific TMA (Figure 3) was made of 98 lung squamous cell carci-nomas that were diagnosed and treated from 1987 to 1992 at the Verona City

Abbreviations: DDR, DNA damage response; DSB, double-strand break;HGD, high-grade dysplasia; IHC, immunohistochemistry; LGD, low-gradedysplasia; SCC, squamous cell carcinoma; TMA, tissue microarray.

yThese authors contributed equally to this work.

� The Author 2007. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected] 2082

Downloaded from https://academic.oup.com/carcin/article-abstract/28/10/2082/2476212by gueston 30 March 2018

Major Hospital (Italy). Criteria to be enrolled into the study included p-stage I(pT1, n 5 49; pT2, n 5 49) according to the 2003 edition of the TNM stagingsystem (a shorthand method incorporating the size of the untreated primarycancer (T), the regional lymph node involvement (N), and/or distant metastasis(M)), radical surgery with extensive mediastinal lymph node dissection toensure accurate staging, no (neo)-adjuvant therapy, minimum 30-day post-operative survival and minimum follow-up of 5 years. Eight mild to moderategrade bronchial dysplastic lesions defined as ‘low-grade dysplasia’ (LGD) andten severe dysplasias/carcinomas in situ defined as ‘high-grade dysplasia’(HGD) adjacent to the tumors were also arrayed together with 19 normalbronchial epithelium counterparts.

Reagents and scoring system used for IHC

Formalin-fixed and paraffin-embedded tissue sections were deparaffinized, re-hydrated, unmasked for 50 minutes in 0.25 mM ethylenediaminetetraaceticacid at 95�C and treated for 5 min with 3% H2O2. Slides were then incubatedovernight at þ4�C with different antibodies, revealed using the EnVision plus/horseradish peroxidase detection system (Dako, Carpinteria, CA) and counter-stained with Hematoxilin. Antibodies used were anti-c-H2AX (1:200, UpstateBiotechnology, Lake Placid, NY), anti-ATM pS1981 (1:400, Rockland Immu-nochemicals Gilbertsville, PA), anti-CHK2 pT68 (1:100, Cell Signaling Tech-nology, Danvers, MA), anti-p53 (1:200, DO1, FIRC Institute of MolecularOncology Foundation monoclonal antibodies facility), anti-caspase 3 active(1:5000, R&D Systems, Minneapolis, MN), anti-Ki67 (1:100, Zymed, SouthSan Francisco, CA) and anti-53BP1 (1:15, gift from T.Halazonetis). A semi-quantitative score was used to assess DDR gene expression: 0, no staining orbarely detectable staining in ,10% of cells; 1, weak staining in �10% of cells;2, moderate staining in �10% of cells and 3, strong staining �10% of cells. Forstatistical analysis purposes, scores 2 and 3 were considered to represent anunequivocal positivity. When scoring for 53BP1, the presence of diffuse and/orredistributed signals in �10% of cells was considered positivity, whereas thecomplete absence of signal was considered negativity. A cut-off of �30% ofpositive cells for p53 and Ki67 was used to establish protein accumulation andhigh proliferation, respectively.

Real-time quantitative reverse transcription–polymerase chain reaction

Total RNA was isolated from frozen tissues (one normal breast, four breastcarcinomas, four lung carcinomas and two stomach carcinomas) with Triazol

method (Invitrogen, Corporation, Carlsbad, CA). One microgram of total RNAwas reverse transcribed (PowerScriptRT, Clontech, Laboratories Inc., Moun-tain View, CA) with 100 ng random primers (Invitrogen). Preparations withoutthe enzyme were used as negative controls. Real-time polymerase chain re-action was carried out for each sample in triplicate on the 7900 HT Fast Real-Time Polymerase Chain Reaction System (Applied Biosystems) using theTAQMAN Assay on Demand specific for 53BP1 (HS00996818_M1) in a finalvolume of 15 ll with the Eurogentech Master mix 2�. The reaction was carriedout for 2 min at 50�C, and then a step of 10 min at 95�C, followed by 40 cyclesof 15 s at 95�C and 60 s at 60�C. To normalize the amount of total RNA presentin each reaction, we amplified the housekeeping gene 18S. All the analyzedtumors were expressed as N-fold 53BP1 mRNA relative to a normal breasttissue sample used as calibrator.

Statistical analysis

Fisher#s exact test was used to calculate statistical significant differences inprotein expression between normal (N) and tumor (T) counterparts (Table I)and between the different stages (normal, LGD, HGD, pT1 and pT2) of lungsquamous cell neoplastic disease (Figure 3). Relationship between genes andcorrelation of each gene with clinical, histological and biological parameters inlung squamous cell carcinoma (SCC) (Table II and data not shown) wereassessed using the Pearson’s Chi-square. Survival analysis was performedusing Kaplan–Meier method and curves were compared by the log-rank test(data not shown). All throughout our analyses, differences were judged signif-icant at confidence levels �95% (P � 0.05).

Results

Antibody validation on fixed material

DDR activation in mammalian cells consists in the nuclear activationby phosphorylation of a set of DDR proteins that are cytologicallydetectable by immunostaining, often in the form of nuclear foci.Although the immunological reagents commonly used for these stud-ies have been extensively validated in vitro on cultured cells, their usein complex tissue specimens has been limited so far. We thereforevalidated a set of these reagents by testing their ability to specifically

Fig. 1. Normal human healthy skin before and after ex vivo exposure to X-rays. Markers of an activated DDR as detected by antibodies raised against histoneH2AX pS139 (c-H2AX), ATM pS1981, 53BP1, CHK2 pT68 and p53 become apparent after treatment in both keratinocytes and stromal cells. Magnifications �40and �60 of the same sample are shown.

Complex engagement of DNA damage response pathways

2083


detect a DDR by IHC on formalin-fixed, paraffin-embedded normalhuman skin, a part of which had previously been X-rays irradiatedex vivo. Figure 1 shows that antibodies raised against phosphorylatedS139 of histone H2AX (c-H2AX) (25), autophosphorylated and there-fore activated ATM pS1981 (8), activated CHK2 pT68 (26) and p53(27), all detect a strong nuclear staining exclusively in the irradiatedsample that is absent in the untreated portion. An antibody against53BP1 generates a diffuse nuclear staining in the non-irradiated sam-ple that becomes stronger and punctuated upon irradiation. A weakATM pS1981 cytoplasmic staining was also occasionally observed,but it was not further considered in this analysis. Overall, the observednuclear-staining mirrors that previously reported in cultured cells andwe concluded that these are reliable reagents that specifically detectthe activation of distinct components of the DDR in archival materialby IHC.

A wide survey of DDR in human tumors of different histotypes

We screened a total of 383 human tumors of different common his-totypes and 147 paired normal tissues using multitumor TMAs (seeMaterials and methods for composition) to assess the engagement ofthe DDR machinery in tumors and in their normal counterparts, whenavailable. TMA is a reliable and representative technology that allowsthe in situ analysis of tumoral and normal tissue samples in a high-throughput and internally controlled format. In this initial screen, wechose to interrogate the DDR-signaling cascade at three nodal points:ATM pS1981, the activated form of a key upstream regulator thatcontrols most DDR events; 53BP1, a central mediator of the DDRand c-H2AX, a downstream chromatin modification event stronglylinked to the generation of DSBs. Results are summarized in Table Iand representative pictures are shown in Figure 2A.

The extent of ATM pS1981 accumulation varied both in normal andtumor samples, depending on the tissue and cell type. ATM activationwas absent or low (�20% of positive samples) in all normal tissuesand in gastrointestinal, kidney and lung cancers (SCC and large cellcarcinomas subtypes), whereas higher levels of ATM pS1981 wereobserved in the other tumor histotypes being significantly up-regulated in breast and larynx cancers compared with their normalcounterparts (Fisher’s exact test, P 5 0.04 and P 5 0.03, respec-tively). The same TMAs were scored for the presence or absence of53BP1. While the majority of normal tissue samples stained positivefor it (with a diffuse nuclear-staining pattern reminiscent of the non-active form), we observed a significant dramatic loss of 53BP1 signalin many tumor types such as lung (P , 0.001), kidney (P 5 0.002),larynx (P 5 0.007) and stomach (P 5 0.004) carcinomas. A subset ofthe same tumors analyzed by IHC–TMA (n 5 10, six 53BP1 IHCpositive and four 53BP1 negative) was subsequently analyzed byquantitative polymerase chain reaction to assess if 53BP1 protein losscorrelates with its transcript. 53BP1 mRNA was indeed generallylower in samples scored negative in IHC than in positive ones (Figure2B). This suggests that altered 53BP1 accumulation in tumors can bethe result of transcriptional modulation. In addition, in one-third oflarynx carcinomas, 53BP1 loss was achieved by nuclear exclusion andcytoplasmic retention. Importantly, all tumors retaining nuclear53BP1 displayed a focal nuclear pattern of staining, highly suggestiveof its activation (Figures 2 and 3).

Similar to other DDR genes, the study of c-H2AX-positive cellsrevealed a tissue type-specific pattern of accumulation. With the onlyexception of lung, no c-H2AX accumulation was found in normalcells. In contrast, half of all kidney tumors stained positive, (T versusN, P 5 0.007) and nearly one non-small cell lung carcinoma in four

Fig. 2. DDR in different tumor histotypes and in non-neoplastic lung. (A) Representative images of normal (N) and paired neoplastic tissues (T) are shown. Forosteogenic sarcomas (OGS) only the tumor tissue is shown. (B) 53BP1 mRNA expressed as fold change (y-axis) relative to expression in normal breast set to 1.Full and empty squares indicate samples scored positive and negative in IHC, respectively. Samples 1, 5, 6 and 8 are lung carcinoma; samples 2, 3, 4 and 7 arebreast carcinoma and samples 9 and 10 are stomach carcinoma. Bar indicates standard error. (C) Activation of DDR was observed in a minority of normalbronchial cells flanking tumors and in non-neoplastic lung only near inflammation but not in the adjacent, non-inflamed bronchial epithelium. Arrows indicateneutrophil granulocytes with multi-lobed nucleus.

P.G.Nuciforo et al.

2084


stained c-H2AX positive—the percentage of positive cases varyingamong different histotypes and it was significantly higher in SCCsfollowed by large cell carcinomas and adenocarcinomas (Table I), inagreement with the reported different genomic instability (28,29). Ouranalysis also revealed that sarcomas, osteogenic sarcomas in particular,were characterized by high levels of both c-H2AX and ATM pS1981 (7out of 10 strongly positive osteogenic sarcomas) and presence of53BP1. Nevertheless, tumors from the breast, larynx and the gastroin-testinal tract very rarely showed c-H2AX accumulation. Notably, inbreast tumors and glioblastomas, we clearly observed ATM activationin the absence of c-H2AX. This intriguing observation suggests thatother events distinct from DSB generation, possibly related to chromatinchanges (8), may contribute to ATM activation in cancer.

As mentioned above, a limited number of samples from normaltissues flanking the tumors stained positive for c-H2AX in lung andfor the activated form of ATM in lung, kidney and the gastrointestinaltract. To assess whether DDR activation in normal tissues was anindirect consequence of the tumor condition, we stained a separateset of large portions of normal lung samples from non-neoplasticpatients (see Materials and methods) for c-H2AX. We discovered thatalso in these samples c-H2AX-positive bronchial cells were clearlydetectable but only near acutely inflamed areas defined by a patholo-gist according to morphological criteria such as the presence of neu-trophil granulocytes in the bronchial epithelium (Figure 2C). These

novel observations (and others not shown) suggest that markers of anactivated DDR, such as c-H2AX, accumulate also in acutely inflamedtissues, therefore, in the absence of frank pre-neoplastic and neoplas-tic condition.

Overall, the screen of multitumor TMA revealed that althoughtumorigenesis can be associated with the activation of the factorsstudied, ATM, 53BP1 and c-H2AX, the extent of their engagementis strongly affected by the organ and the cell types involved. All butone (colon) tumor types showed at least one of the three DDR markersstudied differentially regulated between invasive tumors and normaltissues. Indeed, some tumors lack bona fide markers of DSB accumu-lation such as c-H2AX. Instead, loss of 53BP1 is the most reproduc-ible observation among different cancers analyzed. Furthermore,unanticipatedly, we observed that morphologically normal tissuesflanking a tumor or in association with an ongoing inflammatory pro-cess contain activated components of the DDR apparatus.

DDR activation and suppression during lung squamous cellneoplastic progression

The complex relation between tumor formation and its impact on theDDR cascade at different levels, together with the observed depen-dency on the tumor histotype and the potentially dynamic engagementof DDR signaling, prompted us to study a larger but more homoge-neous set of samples of the same histotype collected at different

Fig. 3. DDR in lung SCC progression. (A) Percentage of positive samples is shown on the vertical axis. The horizontal axis describes the stage of the lesion.Normal lung includes also bronchial respiratory epithelia from multitumor TMA. DDR gene activation is shown on the left line-graph; p53, ki67 and caspase 3activation are on the central line-graph. Statistical differences among groups are calculated with Fisher’s exact test and shown in the right table; Blank cellsindicate lack of statistical significance. N, normal bronchial epithelium. (B) Representative images of lung-specific TMA sections stained for c-H2AX, ATMpS1981, 53BP1, CHK2 pT68 and p53. Normal (N), dysplastic (D, including both LGD and HGD), pT1 and pT2 tumor areas are shown. (C) High-magnificationimages of 53BP1 nuclear pattern in normal lung (N), dysplasia (D) and invasive tumor (T): note 53BP1 redistribution in foci in transformed cells (as shown byarrows) compared with normal epithelium.


2085


disease stages with a broader range of DDR markers. We thereforefollowed DDR engagement during lung squamous cell tumorigenesisin normal (N), LGD and HGD, pT1 and pT2 stage I SCC tissues frompatients with long-term follow-up. In addition to c-H2AX, ATMpS1981 and 53BP1, we analyzed these specimens for the presenceof the activated forms of CHK2 (CHK2 pT68), together with caspase3 active, p53 and Ki67.

The results of this analysis, summarized in Figure 3, strengthenedthose obtained in the first screening and increased its resolution. Con-sistent with our previous set of samples, we observed a progressiveand dramatic loss of detectable 53BP1 during tumor progression(Figure 3A, left panel; right panel for P values) from as early as theN to LGD transition. ATM pS1981 and CHK2 pT68 are present in asmall percentage of normal bronchial cells and they progressively

increase up to HGD (N versus HGD, P values 5 0.05 for ATMpS1981 and 0.01 for CHK2 pT68) to dramatically drop in pT2(HGD versus pT2, P values 5 0.02 for ATM pS1981 and 0.01 forCHK2 pT68). Differently, although a slight, non-significant, increasein c-H2AX positivity was observed from normal to pre-neoplastic/neoplastic condition, c-H2AX levels remain similar in all samplesanalyzed, independently from the disease stage. Tumor progressionwas associated with an initial accumulation of activated caspase 3(from N to LGD, P 5 0.009) and a subsequent loss, coincident withp53 inactivation by mutation, as suggested by its increased stability(N versus HGD, P 5 0.0001) (30). Overall, stepwise 53BP1 loss, p53mutation and ATM/CHK2 inactivation lead to a progressive increaseof Ki67-positive cells (Figure 3A, central panel; right panel forP values) and it is consistent with a tumor suppressive role of DDRcomponents.

Analysis of activated DDR pathways in SCC tumors and theircorrelations with clinicopathologic parameters

Despite a wealth of knowledge has been gained on the mechanisms ofDDR activation and the hierarchy of its components in in vitro studies,very little is known about the same events in the context of in vivohuman tumorigenesis. We therefore analyzed the correlations amongactivated DDR components in individual tumor samples (pT1 andpT2) (Table II). Among all statistically significant correlations, thepresence of activated ATM was most likely detected in tumors main-taining 53BP1 expression and the T68 phosphorylated form of CHK2.Since ATM, CHK2 and 53BP1 activations are known to be intimatelylinked and part of the same pathway, these results validate our anal-ysis and confirm previously published data utilizing human in vitrosystems and mouse genetics (10,11,31). Unexpectedly, c-H2AX didnot correlate with phosphorylated ATM. This suggests that its main-tenance may rely on the activity of different kinases, such as AtaxiaTelangiectasia-Related (ATR) or DNA-activated protein (DNA-PK).Interestingly, accumulation of p53 protein, apoptosis (as determinedby caspase 3 activation) and proliferation markers (Ki67) did notsignificantly correlate with the activation of any individual DDR fac-tor in invasive tumors. We interpret this as suggestive that activationof a single DDR component or linear pathway in tumors is not suffi-cient to induce apoptosis but, instead, the choice to undergo cellsuicide or proliferation can be the outcome only of multiple feedsfrom different DDR components and pathways.

Such a homogeneous set of samples allowed us to search for cor-relations between activated components of the DDR-signaling path-ways and clinicopathologic parameters. We discovered a significantcorrelation between reduced local tumor extension (as assessed byTNM) and DDR activation as detected by phosphorylation of ATM(P 5 0.05), CHK2 (P 5 0.02) and the presence of 53BP1 (P 5 0.01),strengthening the notion of a tumor suppressive role of DDR activi-ties. However, there is no significant association between individualactivated DDR markers and age, smoking habits, relapse, vital status,overall and disease-free survival (data not shown).

Discussion

Here, we have shown that several key components of the DDRare activated in a broad collection of human tumors of different

Table I. Multitumor TMA analysis of DDR activation

Organ Histotype

ATM pS1981 53BP1 c-H2AX

Positive (n) Positive (n) Positive (n)

Lung Normal 11% (63) 86% (64) 13% (52)SCC 15% (59) 27% (62)� 33% (60)�AC 23% (34) 33% (36)� 6% (32)LCC 0% (9) 33% (9)� 22% (9)

Breast N/FC 0 (10) 91% (11) 0 (8)FA 48% (23)� 100% (23) 0 (27)IDC 36% (39)� 74% (39) 5% (43)

Colon–rectum Normal 14% (7) 85% (13) 0 (9)Carcinoma 20% (5) 63% (16) 0 (18)

Kidney Normal 13% (15) 87% (15) 0 (13)CCC 11% (18) 29% (17)� 47% (15)�

Larynx Normal 0 (17) 70% (20) 0 (16)SCC 25% (24)� 28% (25)� 7% (28)

Stomach Normal 9% (23) 62% (24) 0 (10)AC 0 (27) 21% (28)� 0 (11)

Hematopoietic NHL 29% (17) 55% (20) 0 (19)HL 25% (4) 75% (8) 0 (8)

Skin Melanoma 39% (18) 68% (19) 0 (20)Nevus 36% (11) 64% (14) 0 (14)

Soft tissue STT 33% (18) 90% (10) 22% (18)

Bone OGS 70% (10) 100% (10) 70% (10)

CNS Meningioma 50% (4) 25% (4) 25% (4)

GBL 68% (19) 100% (20) 0 (20)

Data are shown by organ of origin and histotype. The percentage of positivesamples next to the total number of samples studied is shown—see Materialsand methods for scoring criteria. In the lung, the normal tissue wasrepresented by both bronchial and alveolar epithelium. N/FC, normal breast/fibrocystic disease; FA, fibroadenoma; IDC, infiltrating ductal carcinoma;CCC, clear cell carcinoma; AC, adenocarcinoma; LCC, large cell carcinoma;NHL, non-Hodgkin’s lymphoma; HL, Hodgkin’s lymphoma; STT, soft tissuetumor and OGS, osteogenic sarcoma. Asterisks indicate statisticallysignificant differences between normal and tumor samples as assessed byFisher’s exact test.

Table II. Statistical correlations among individual activated DDR components and cell-cycle markers as measured by a Pearson’s Chi-square test

c-H2AX ATM pS1981 53BP1 CHK2 pT68 p53 Caspase 3 act

ATM pS1981 0.253BP1 0.1 ,0.0001CHK2 pT68 0.02 0.008 0.0002p53 0.9 0.6 0.7 0.7Caspase 3 act 0.6 0.2 0.6 0.2 0.3Ki67 0.9 0.8 0.9 0.8 0.1 0.3

Positive correlations are indicated in italic.

P.G.Nuciforo et al.

2086


histotypes. Our results are consistent with the emerging evidence thatDDR components are misregulated in tumorigenesis (16–22). Never-theless, differently from published reports, our data demonstrate thatthe scenario is complex and suggest a word of caution before drawinga univocal and exclusive correlation between tumorigenesis and fullDDR activation. In fact, we observed that individual DDR factors areactivated or, in the case of 53BP1, lost with different frequenciesaccording to the tumor histotypes. Although DDR in tumorigenesisis a dynamic process and therefore the extent of the detectable DDRactivation depends on the stage of the tumor, our data are consistentwith a potentially dissimilar ability of different oncogenic events topreferentially activate/inactivate individual DDR pathways to variableextents. The reason for the observed differences may also lie in thereported different ability of some cell types to mount a DDR (32),possibly also in relation to the different subcellular localization ofsome DDR factors (19,33,34) and the ability of some pathways toimpact on DDR activation (35).

Of note, we detected a DDR in a number of different normal tissuesamples, including a subset of lung samples in which normal bron-chial epithelial cells showed activation of ATM, CHK2 and H2AX.This may reflect the cell response to DNA damage generated byexposure to exogenous agents and endogenous cellular events. Inter-estingly, we found that acute inflammatory processes may be associ-ated with a DDR. This is intriguing, because a link betweeninflammation and transformation has been proposed (36) and our data,together with reports of increased loss of heterozygosity in normalepithelium next to lung carcinomas (28) and in inflamed tissues(28,37), support the hypothesis that a non-oncogenic or pre-oncogenicprocess, such as inflammation, may engage components of the DDRmachinery. In this scenario, the selective pressure for the loss of DDRmay stimulate the neoplastic growth. However, it is also possible thatthe observed DDR activation in normal tissues next to tumors mayresult from a bystander effect (38).

Our vertical analysis of lung SCC highlights the dynamic and com-plex involvement of DDR-signaling pathway. 53BP1 loss is the ear-liest event observed, occurring in the transition from normal to LGD;this observation suggests a pivotal sentinel role of this DDR factor inlung tumor progression. Differently, ATM and CHK2 activations peakin HGD and they coincide with p53 inactivation. As expected, loss ofp53 functions reduce the fraction of apoptotic cells, as assessed bya decrease in active caspase 3. Later, cancer progression toward franktumor formation is associated with progressive loss of the activatedforms of ATM and CHK2, with the most significant reduction at thepT1–pT2 step. Differently, c-H2AX does not seem to be significantlymodulated among the samples studied.

Therefore, 53BP1 loss, p53 mutation and ATM/CHK2 inactivationare three stage-specific events that, once integrated, provide a concep-tual framework to explain the otherwise progressive increase in ki67-positive, and therefore proliferating, cells and tumor evolution.

These data suggest that transformation exerts a tremendous selec-tive pressure against the accumulation of these proteins and stronglypoint to a tumor suppressive function of DDR components, 53BP1 inparticular. This is consistent with the lymphoma-prone phenotype of53BP1 knockout mice (39). Although the molecular mechanisms of53BP1 tumor suppressive functions are not fully known, its inactiva-tion leads to escape from the proliferative block associated withoncogene-induced cellular senescence (10) and to the impairmentof p53-dependent apoptosis (13).

Our results may also have potentially important consequences forthe use of small molecule inhibitors of ‘druggable’ DDR components(40). These agents are pursued with the idea of sensitizing tumors totherapy or in the hope that the loss of a DNA repair pathway mayrender cancer cells more susceptible to the inactivation of a parallelpathway (41,42). Our results may help choosing tumor histotypes thatpreferentially activate, or not, a given DDR pathway and may there-fore respond differently to this approach. Furthermore, as the activa-tion of many DDR proteins is lost during tumor progression, the stageof the tumor should also be considered. Finally, in the light of thetumor suppressive functions of some DDR factors, the use of DDR

inhibitors in the absence of additional DNA-damaging agents may, infact, accelerate tumor growth. Therefore, our results suggest that theuse of DDR modulators should greatly benefit from a careful analysisof the DDR activation pattern of the individual tumor before treat-ment. Furthermore, since we noticed that some DDR components areactivated in normal tissues, DDR inhibitors may have unexpected sideeffects in non-neoplastic tissues.

Funding

Associazione Italiana per la Ricerca sul Cancro (AIRC) to F.d.A.d.F.

Acknowledgements

We thank Pier Paolo Di Fiore for support and critical reading of the manuscript.T. Halazonetis and FIRC Institute of Molecular Oncology Foundation mono-clonal facility for reagents.

Conflict of Interest statement: None declared.

References

1.Shiloh,Y. (2006) The ATM-mediated DNA-damage response: taking shape.Trends Biochem Sci, 31, 402–410.

2. Jazayeri,A. et al. (2006) ATM- and cell cycle-dependent regulation of ATRin response to DNA double-strand breaks. Nat. Cell Biol., 8, 37–45.

3.Stucki,M. et al. (2005) MDC1 directly binds phosphorylated histone H2AXto regulate cellular responses to DNA double-strand breaks. Cell, 123,1213–1226.

4.Huyen,Y. et al. (2004) Methylated lysine 79 of histone H3 targets 53BP1 toDNA double-strand breaks. Nature, 432, 406–411.

5.d’Adda di Fagagna,F. et al. (2003) A DNA damage checkpoint response intelomere-initiated senescence. Nature, 426, 194–198.

6.Herbig,U. et al. (2004) Telomere shortening triggers senescence of humancells through a pathway involving ATM, p53, and p21(CIP1), but notp16(INK4a). Mol. Cell, 14, 501–513.

7.Behrend,L. et al. (2003) Reactive oxygen species in oncogenic transforma-tion. Biochem. Soc. Trans., 31, 1441–1444.

8.Bakkenist,C.J. et al. (2003) DNA damage activates ATM through intermo-lecular autophosphorylation and dimer dissociation. Nature, 421, 499–506.

9.Gibson,S.L. et al. (2005) Hypoxia-induced phosphorylation of Chk2 in anAtaxia telangiectasia mutated-dependent manner. Cancer Res., 65, 10734–10741.

10.Di Micco,R. et al. (2006) Oncogene-induced senescence is a DNA damageresponse triggered by DNA hyper-replication. Nature, 444, 638–642.

11.Bartkova,J. et al. (2006) Oncogene-induced senescence is part of the tu-morigenesis barrier imposed by DNA damage checkpoints. Nature, 444,633–637.

12.Bartkova,J. et al. (2005) DNA damage response as a candidate anti-cancerbarrier in early human tumorigenesis. Nature, 434, 864–870.

13.Gorgoulis,V.G. et al. (2005) Activation of the DNA damage checkpoint andgenomic instability in human precancerous lesions. Nature, 434, 907–913.

14.Kastan,M.B. et al. (2004) Cell-cycle checkpoints and cancer. Nature, 432,316–323.

15.O’Driscoll,M. et al. (2006) The role of double-strand break repair—insights from human genetics, Nat Rev Gentet, 7, 45–54.

16.Bartkova,J. et al. (2004) Aberrations of the Chk2 tumour suppressor inadvanced urinary bladder cancer. Oncogene, 23, 8545–8551.

17.Staalesen,V. et al. (2004) Alternative splicing and mutation status ofCHEK2 in stage III breast cancer. Oncogene, 23, 8535–8544.

18.Angele,S. et al. (2003) Altered expression of DNA double-strand breakdetection and repair proteins in breast carcinomas. Histopathology, 43,347–353.

19.Kairouz,R. et al. (1999) ATM protein synthesis patterns in sporadic breastcancer. Mol. Pathol., 52, 252–256.

20.Wilson,C.A. et al. (1999) Localization of human BRCA1 and its loss inhigh-grade, non-inherited breast carcinomas. Nat. Genet., 21, 236–240.

21.Vo,Q.N. et al. (2004) The ATM gene is a target for epigenetic silencing inlocally advanced breast cancer. Oncogene, 23, 9432–9437.

22.Ding,S.L. et al. (2004) Abnormality of the DNA double-strand-breakcheckpoint/repair genes, ATM, BRCA1 and TP53, in breast cancer is re-lated to tumour grade. Br. J. Cancer, 90, 1995–2001.


2087


23.Zhang,P. et al. (2004) CHK2 kinase expression is down-regulated due topromoter methylation in non-small cell lung cancer. Mol. Cancer, 3, 14.

24.Kononen,J. et al. (1998) Tissue microarrays for high-throughput molecularprofiling of tumor specimens. Nat. Med., 4, 844–847.

25.Rogakou,E.P. et al. (1999) Megabase chromatin domains involved in DNAdouble-strand breaks in vivo. J. Cell Biol., 146, 905–916.

26.Melchionna,R. et al. (2000) Threonine 68 is required for radiation-induced phosphorylation and activation of Cds1. Nat. Cell Biol., 2, 762–765.

27.Levine,A.J. (1997) p53, the cellular gatekeeper for growth and division.Cell, 88, 323–331.

28.Wistuba,I.I. et al. (1999) Sequential molecular abnormalities are involvedin the multistage development of squamous cell lung carcinoma. Oncogene,18, 643–650.

29.Wistuba,I.I. et al. (2000) High Resolution Chromosome 3p Allelotyping ofHuman Lung Cancer and Preneoplastic/Preinvasive Bronchial EpitheliumReveals Multiple, Discontinuous Sites of 3p Allele Loss and Three Regionsof Frequent Breakpoints. Cancer Res., 60, 1949–1960.

30.Hall,P.A. et al. (1994) p53 in tumour pathology: can we trust immunohis-tochemistry?–Revisited!. J. Pathol., 172, 1–4.

31.Ahn,J. et al. (2004) The Chk2 protein kinase. DNA Repair (Amst.), 3, 1039–1047.

32.Dmitrieva,N.I. et al. (2004) Cells adapted to high NaCl have many DNAbreaks and impaired DNA repair both in cell culture and in vivo. Proc. NatlAcad. Sci. USA, 101, 2317–2322.

33.Lukas,C. et al. (2001) DNA damage-activated kinase Chk2 is independentof proliferation or differentiation yet correlates with tissue biology. CancerRes., 61, 4990–4993.

34.Barlow,C. et al. (2000) ATM is a cytoplasmic protein in mouse brain requiredto prevent lysosomal accumulation. Proc. Natl Acad. Sci. USA, 97, 871–876.

35.Satyanarayana,A. et al. (2004) Mitogen stimulation cooperates with telo-mere shortening to activate DNA damage responses and senescence signal-ing. Mol. Cell. Biol., 24, 5459–5474.

36.Balkwill,F. et al. (2005) Smoldering and polarized inflammation in theinitiation and promotion of malignant disease. Cancer Cell, 7, 211–217.

37.Park,W.S. et al. (1998) Loss of heterozygosity and microsatellite instabilityin non-neoplastic mucosa from patients with chronic ulcerative colitis. Int.J. Mol. Med., 2, 221–224.

38.Sokolov,M.V. et al. (2005) Ionizing radiation induces DNA double-strandbreaks in bystander primary human fibroblasts. Oncogene, 24, 7257–7265.

39.Ward,I.M. et al. (2003) p53 binding protein 53BP1 is required for DNA dam-age responses and tumor suppression in mice. Mol. Cell. Biol., 23, 2556–2563.

40.Zhou,B.B. et al. (2004) Targeting the checkpoint kinases: chemosensitiza-tion versus chemoprotection. Nat. Rev. Cancer, 4, 216–225.

41.Bryant,H.E. et al. (2005) Specific killing of BRCA2-deficient tumours withinhibitors of poly(ADP-ribose) polymerase. Nature, 434, 913–917.

42.Farmer,H. et al. (2005) Targeting the DNA repair defect in BRCA mutantcells as a therapeutic strategy. Nature, 434, 917–921.

Received December 29, 2006; revised April 27, 2007; accepted April 30, 2007

P.G.Nuciforo et al.

2088


Documents

Complex engagement of DNA-damage response pathways in