Overexpression of chromatin assembly factor-1 p60, poly(ADP-ribose) polymerase 1 and nestin predicts metastasizing behaviour of oral cancer

Aims The natural history of oral squamous cell carcinomas (OSCCs) is variable and difficult to predict. This study aimed to assess the value of the expression of poly(ADP-ribose) polymerase 1 (PARP-1), chromatin assembly factor-1 (CAF-1)/p60 and the stem cell markers CD133, CD166, CD44, CD44v6 and nestin as markers of outcome and progression-free survival in OSCC patients. Methods Clinical data were collected from 66 patients (41 male and 25 female, aged 29–92 years) who underwent surgery for OSCC of the tongue, floor, lips, and palate. During follow-up (range: 12–131 months), 14 patients experienced relapse/metastasis and/or death. The study was performed by immunohistochemistry on paraffin-embedded tumour tissues, western blot analysis of tumour protein lysates and human cell lines, and RNA silencing assays. In addition, the human papillomavirus (HPV) status of primary tumours was evaluated by immunohistochemistry and viral subtyping. Univariate and multivariate analyses were performed to determine the correlation between these parameters and the clinical and pathological variables of the study population. Results and conclusions We found that a PARP-1high/CAF-1 p60high/nestinhigh phenotype characterized the OSCCs with the worst prognosis (all HPV-negative). This may be of benefit in clinical management, since radio-enhancing anti-PARP-1 and/or anti-CAF-1/p60 agents may allow radioresistance to be bypassed in the nestin-overexpressing, metastasizing OSCC cells.


Introduction
Oral squamous cell carcinoma (OSCC) is the most frequent cancer of the head and neck region. It represents the fifth most common cancer worldwide, and is the sixth most common cause of cancer-related deaths per year in the USA. 1,2 The major risk factors for this tumour are prolonged exposure to tobacco and alcohol. In many cases surgery leaves patients disfigured and ⁄ or necessitates further reconstructive procedures. 1,3 Unfortunately, OSCCs are frequently unresponsive to alternative therapeutic options, such as radiotherapy and chemotherapy, which target the hierarchically organized, rapidly dividing tumour cells constituting the bulk of the tumour mass. 4,5 It has been hypothesized that subsets of self-renewing, proliferatively quiescent tumour cells with stem-like properties, known as cancer stem cells (CSCs), may account for the resistance to DNAdamaging agents and the failure of long-term disease control of malignant tumours. [6][7][8][9][10][11][12][13][14][15] The defence strategies of cancer cells against the cytotoxic damage and apoptotic DNA double-strand breaks induced by therapeutic ionizing radiation or alkylating agents encompass the rapid synthesis and degradation of poly(ADP-ribose) by cellular poly(ADP-ribose) polymerases (PARPs). [16][17][18] The addition of poly(ADP-ribose) groups to histones induces nucleosome modification and relaxation of the chromatin structure, facilitating DNA repair through the action of large protein complexes, such as the molecular chaperone chromatin assembly factor-1 (CAF-1). [18][19][20][21] CAF-1 comprises a complex of three subunits (p150, p60, and p48) and drives the incorporation and assembly of H3K56-acetylated histones into chromatin in response to oxidative stress, DNA damage, and mismatch-containing strands, restoring chromatin structure on the completion of double-strand break repair. [22][23][24][25][26][27] Knocking down p60 disrupts the activity of CAF-1 in nucleosome assembly, causing replication fork arrest, activation of the intra-S-phase checkpoint, and global defects in chromatin structure. [28][29][30][31] The nuclear expression of CAF-1 ⁄ p60 is increased in multiple types of cancer, proportionally to their adverse clinical behaviour. 23,[31][32][33][34] On account of these postulates, we examined the expression of the stem cell-associated antigens CD133, CD166, CD44 with the v6 variant and nestin (frequently associated with CSCs of solid malignancies) in a series of primary and metastatic OSCCs. 35 We then compared these results with the expression levels of PARP-1 and CAF-1 ⁄ p60 proteins in the same group of tumours. In addition, on the basis of evidence that a distinct cohort of head and neck cancers testing positive for high-risk (HR) human papillomavirus (HPV) DNA shows less aggressive behaviour and better response to therapy, we related our data to the HPV status of primary tumours. [36][37][38][39] As HR HPVs cause strong expression of p16 INK4a , a component of the Rb tumour suppressor gene pathway, and immunohistochemistry for p16 INK4a is considered to be a reasonable surrogate for detecting transcriptionally active HPV infection, p16 INK4a expression was also investigated in our cases of OSCC. [40][41][42][43] Finally, we explored the existence of any significant association between the several parameters that we analysed and the clinical course of OSCC.

Materials and methods patients and tissue samples
Formalin-fixed, paraffin-embedded tissue blocks of 66 primary OSCCs (41 tongue, 16 floor, five lip and four palate squamous cell carcinomas) and corresponding metastases, diagnosed and excised with healthy surgical margins from January 2000 to December 2009, were retrieved from the archives of the Pathology Section of the Department of Biomorphological and Functional Sciences, 'Federico II' University of Naples. The clinical data and pathological features of the tumours are reported in Table 1. 44 No patient experienced radiotherapy before surgery. The study design and procedures involving tissue samples collection and handling were performed according to the Declaration of Helsinki, in agreement with the current Italian law, and to the Institutional Ethical Committee guidelines.

tissue microarray ( tma) construction
The most representative tumour area for each case was selected on a donor tissue block under the guidance of the corresponding haematoxylin and eosin (H&E) section. A manual tissue microarrayer (Tissue-Tek Quick-Ray Sakura, Torrance, CA, USA) was used to punch one cylindrical core tissue specimen (3 mm diameter) from the selected area. The cores were implanted into a recipient paraffin block to construct the TMA, each containing two samples of non-neoplastic oral mucosa (control). After chilling at )10°C for 30 min, 43 several 4-lm sections were cut from each TMA. The first section was stained with H&E to confirm both the presence of the selected areas from each tumour and the integrity of tissues. The other sections were mounted on poly-lysine coated glass slides.

cell culture
Normal human keratinocytes (HNEK) were cultured in keratinocyte growth medium (Cambrex, East Rutherford, NJ, USA). BHY, CAL33 and HN cell lines are described elsewhere. 53 HN and BHY cell lines were derived from a human oral cavity SCC, and CAL33 cells from human tongue SCC. Cells were maintained in DMEM supplemented with 10% fetal bovine serum, 2 mm l-glutamine and 100 units ⁄ ml penicillin-  was considered to be significant if it was <0.05. The statistical analysis was performed using the r statistical package v. 2.10.1 (R Foundation for Statistical Computing, Vienna, Austria). Evaluation of the intraobserver and interobserver agreement for the tested proteins on whole sections and TMAs and of the concordance between sections and TMAs was performed by use of Cohen's weighted kappa statistic. 55 Results immunohistochemical staining of tmas and routine tissue sections All of the antibodies used in this study showed a high concordance level between the expression values on routine sections and those on TMA sections, as reported below. For each antibody, we report the results of the immunohistochemical staining, without specifying whether the data refer to TMA sections or whole sections of tumours. The immunohistochemistry results for each antibody analysed are reported in detail in Table 3. CAF-1 ⁄ p60, PARP-1 and nestin were expressed in all OSCCs; tumours were assigned to two (low-score and high-score) categories on the basis of the median values of marker expression ( Table 2). All the tumours showing the simultaneous overexpression of CAF-1 ⁄ p60, PARP-1 and nestin, developed, during followup, recurrence and ⁄ or metastases or death ( Figure 1; Table 3). This overexpression was retained by the corresponding metastatic tissues ( Figure 1). Among the 36 cases showing high-level CD44 expression, only 12 (33%) developed an adverse event during follow-up ( Table 3). The immunohistochemical expression of CD166, CD133 and CD44v6 was variable, and did not show a significant association with clinical behaviour. The p16 INK4a -positive tumours were all positive for HPV16 (Table 3). Among p16 INK4a -positive tumours, CAF-1 ⁄ p60, PARP-1 and nestin were only barely detectable ( Figure 2). In TMAs, as in routine sections, focal immunoreactivity for the evaluated proteins was observed in a few keratinocytes of normal oral epithelium, almost always localized to the basal layer.
protein expression of stem cell markers, parp-1 and caf-1/p 6 0 in oscc and cell lines To verify the immunohistochemical data, we harvested protein lysates from selected high-stage ⁄ grade snap-frozen OSCC samples (n = 10) and from the corresponding normal mucosa, and examined CAF-1 ⁄ p60 protein levels by immunoblotting. We found CAF-1 ⁄ p60 protein to be highly expressed in all carcinomas, particularly in metastatic tissue ( Figure 3). The uncleaved p111 PARP-1 isoform was present in large amounts in tumours and metastases, as compared with normal counterparts (Figure 3). We previously analysed the same group of OSCC primary samples, and found high levels of CD44 and its v6 variant, which was particularly expressed in highgrade tumours. In the present study, these samples were also evaluated for the expression of nestin, CD133, and CD166. These stem cell markers showed greater expression in OSCC than in normal oral mucosa; the highest levels were registered for nestin, particularly in OSCC metastases ( Figure 3). We also evaluated the expression levels of CAF-1 ⁄ p60, PARP-1 and stem cell markers in some cultured human OSCC lines. CAF-1 ⁄ p60 and PARP-1 were expressed at the highest level in OSCC cell lines (Figure 3), which also showed high levels of CD44v6, nestin and CD166 expression (Figure 3).

statistical analysis
The level of agreement for the immunohistochemical staining evaluation, expressed by the kappa coefficient, was >0.75 for both intraobserver and interobserver evaluations, on sections and on TMAs, for all of the antibodies used in the present study. The concordance between the expression levels evaluated on the whole sections and on TMA sections was high (kappa coefficient >0. 75).
No statistical differences for age, gender, tumour subsite, grade or stage of disease were found between patients who had different clinical outcomes (Table S1). CAF-1 ⁄ p60, PARP-1, CD166, nestin, CD44 and CD44v6 were expressed at significantly higher PARP-1, CAF-1 p60 and stem cells in OSCC 1093 levels in patients who had an adverse event during the follow-up. CAF-1 ⁄ p60, PARP-1 and nestin showed the highest sensitivity (1.00) and specificity (0.64, 0.54, and 0.56, respectively) ( Table 4). Similar results were obtained on evaluating these multiple markers against a single adverse event (Tables S2-S4). As expected, the various markers that we investigated were statistically correlated with each other (Tables 5 and 6). We focused on the three proteins (CAF-1 ⁄ p60, PARP-1, and nestin) that showed the strongest correlation with each other (Table 5) and with tumour biological behaviour.
They were significantly associated with adverse events (P < 0.001), showing sensitivity or specificity levels that reached 1.00. The 14 patients who had simultaneous maximum expression of CAF-1 ⁄ p60, PARP-1 and nestin had at least one adverse event, with an overall median event-free time of 12 months. In contrast, the 52 patients with low expression of at least one of the three proteins in their primary tumours had a favourable outcome (Table 7). Nineteen patients showed high CAF-1 ⁄ p60 and low PARP-1 and ⁄ or nestin expression; none of them had any adverse event during the follow-up. Table 8   Table 3. Immunohistochemical staining and HPV genotyping results, tumour staging and follow-up of patients I  II  III  IVA  Follow- summarizes our findings: the simultaneous 'triplehigh expression' of CAF1 ⁄ p60, PARP-1 and nestin correlates, with high sensitivity and specificity, with relapse, metastasis, and death.

Discussion
Our study provides evidence for the existence of a definite inter-relationship between the overexpression

tumour heterogeneity and expression of stem cell markers in osccs
The presence of non-cycling CSCs, which are inherently resistant to chemotherapy and radiotherapy, greatly contributes to the metastasizing ability of cancers. [13][14][15][56][57][58] Unfortunately, CSCs may be heterogeneous from patient to patient and within each tumour type and single tumour clones. 56,[59][60][61][62] Moreover, given that, currently, specific markers are not able to unequivocally distinguish CSCs from their differentiating progeny and their normal tissue counterparts, the goal of specifically targeting CSCs, avoiding unwanted toxicity to normal stem cells, seems far from being reached. 13,63 The metastasizing OSCCs of our study showed high cellular levels of CD133, CD44, CD44v6, and CD166; their coexpression, rather than being specifically clustered in a constant fashion, was randomly distributed. This finding is reminiscent of the variable combination of expression between a stem cell marker (CD44) and other stem cell markers, as recently reported for head and neck squamous cell carcinoma, and fits with the hypothesis that CSC evolution is a continuous dynamic process, which constantly leads to new generations of CSCs with genetic and ⁄ or epigenetic changes favouring metastatic potential. 13,64 Among all of the stem cell markers that PARP-1, CAF-1 p60 and stem cells in OSCC 1099 we evaluated, only nestin was maximally expressed in 100% of metastasizing OSCCs. According to this point of view, the nestin high CSCs should be considered to be the result of the selection of a metastatic clone among the heterogeneous pool of OSCC cells expressing stem cell markers.
parp-1 and caf-1 / p60 expression in osccs PARP-1 and CAF-1 ⁄ p60 nuclear proteins cross-talk with several molecular pathways involved in histone acetylation. The poly(ADP-ribosyl)ation of histones effectively functions like acetylation, maintaining chromatin nucleosomes in a fully relaxed, transcriptionally active state. Either PARP-1 [65][66][67][68] or CAF-1 ⁄ p60 [31][32][33][34] overexpression has been reported in multiple types of cancer, and this has been related to histopathological grade and ⁄ or adverse clinical behaviour. Their expression could help tumour cells to withstand genotoxic stress, by increasing their resistance to DNA-damaging agents, and may result in radoresistance and chemo-resistance. In the OSCCs analysed, the metastasizing group showed striking immunoreactivity for both CAF-1 ⁄ p60 and PARP-1, either on tissue sections or in tumour protein lysates. Accordingly, the protein expression analysis of HN, BHY and CAL33 cells showed increased amounts of CAF-1 ⁄ p60 and PARP-1 in OSCCs with respect to controls. These data are in agreement with previous reports of CAF-1 ⁄ p60 and PARP-1 as markers of adverse biological behaviour in oral and head and neck squamous cell carcinomas. 35,69 Moreover, we found that full-length PARP-1 increased progressively from controls to primary tumours and cancer cell lines, constituting the only detectable PARP-1 form in metastatic tissue. This finding is in line with results from previous studies performed by our team on human skin melanomas, showing the existence of a direct correlation between melanoma cell aggressiveness and overexpression of full-length PARP-1, evaluated by western blot analysis, and with the reported strong reduction of PARP-1 cleavage in cisplatin-resistant melanoma cell lines. 70 As in the case of skin melanoma, our findings suggest that inhibition of the PARP-1-mediated apoptotic process may also be involved in the worst behaviour of a subset of OSCCs, as the cleaved form of PARP-1 is nearly undetectable in metastatic cases. As is common knowledge, cleavage of the stable 85-kDa fragment from full-length PARP-1 is a caspase-mediated early apoptotic event. Conversely, full-length PARP-1 facilitates the survival of proliferating cells under conditions of DNA damage, mainly via DNA base-excision repair, acting as a negative regulator of genome instability in both normal and neoplastic cells. [71][72][73] Bü rkle et al. 74 showed that overexpression of full-length, wild-type   PARP-1 in transfected hamster cells led to a striking cytoprotective effect, by suppression of DNA damageinduced genomic instability in proliferating cells exposed to genotoxic stress. Moreover, a recent metaanalysis carried out in a large retrospective gene expression dataset revealed that both PARP-1 mRNA and PARP-1 evaluated by immunohistochemistry were overexpressed in the subset of breast cancer with the worst prognosis, in terms of metastasis-free survival and overall survival. This led to the conclusion that nuclear PARP-1 overexpression is an independent prognostic factor for disease-free and overall survival of patients. They hypothesized that PARP-1 overexpression, in some cases, may result from defective PARP-1 cleavage, resulting in reduced tumour apoptosis (see also the previous report of Tang et al.). 72,75 Similar results were obtained in a study of ovarian cancer. 76 We have demonstrated, for the first time, that metastasizing OSCCs are characterized by constant high expression of PARP-1 (almost entirely full-length) coupled to hyperexpression of CAF-1 ⁄ p60 and nestin, regardless of the grade and size of tumours at diagnosis.

hpv status and molecular phenotype of osccs
The link between persistent HR HPV infection and the development of preneoplastic and cancerous lesions of human mucosal epithelia is well known. In head and neck cancers, a clear relationship between HR HPV infection and oropharyngeal cancer has been shown, whereas the data concerning the link between HPV and oral cancer are still not conclusive. The reported prevalence of HPV in oral epithelial neoplastic and preneoplastic lesions ranges from 30% to 86%. [77][78][79][80][81][82] This could be attributable to the lack of standardization of the criteria used for population selection among the different studies. In addition, a meta-analysis of 94 studies investigating 4680 OSCCs revealed that a significantly higher frequency of HPV infection was found if PCR techniques were used than when Southern blot, dot blotting and in-situ hybridization were employed. 77 Nevertheless, these findings are in agreement with the hypothesis that HPV may act as an initiator of epithelial proliferation in oral carcinogenesis, independent of the anatomical subsite (tongue, gingiva, cheek, and oral floor). 77,78,81,83 Our data provide useful information about the molecular events underlying the biology of HPV-positive OSCC. 84 The 10 HPV-positive OSCCs of our series had low to moderate expression of CAF-1 ⁄ p60 and PARP-1, and never showed statistically significant immunoreactivity for stem cell markers. This observation may contribute to explaining the high rate of responsiveness to radio-therapy of HPV-positive head and neck and oral squamous cancers. HR HPV E7 oncoprotein binds and degrades Rb protein, leading to an increase in p16 INK4a levels and to deregulated tumour cell proliferation. 43,85 Stem cells normally reside in a hypoxic niche, where self-renewal and differentiation activity are balanced. When cell proliferation becomes a dominant feature, the expansion of progenitor cells can also occur. This could result in stem cell pool exhaustion. 86 The barely detectable level of HPVpositive OSCC cells with a CSC phenotype may then represent the consequence of continuous HPV-induced p16 INK4a overexpression which might, in the long term, cause defects in the maintenance of stem cell selfrenewal ability in these tumours. This hypothesis is in line with the reported long time lag (from 15 to 30 years) between an oral HPV infection and the development of HPV-related OSCC, and with the reported major impact of p16 INK4a expression on response to treatment and overall survival of patients with head and neck cancer treated using conventional radiotherapy. 87,88 Low levels of PARP-1 and CAF-1 ⁄ p60 with the absence of a significant number of HPV-positive tumour cells showing a CSC phenotype could therefore explain the reduced DNA-repair ability and radioresistance that characterizes this subgroup of OSCCs.

prognostic and therapeutic implications
The significant association between the PARP-1 high ⁄ CAF-1 p60 high ⁄ nestin high phenotypes allows us to hypothesize that it could represent an epiphenomenon of the metastasizing CSC (mCSC) compartment of HPV-negative OSCCs. This finding may have relevance for clinical practice. In fact, it is now recognised that inhibition of PARPs leads to impairment of DNA double-strand-break repair, enhancing the cytotoxic effects of ionizing radiation and DNA-damaging chemotherapeutic agents. 16,89,90 In addition, CAF-1 ⁄ p60 has recently emerged as a promising target, inhibition of which could lead to cell death in aggressive tumours. [32][33][34][35][36] This hypothesis has been further supported by the results of the cell death assay on HaCat and CAL33 cells that we performed with CAF-1 p60 siRNA and the PARP-1 inhibitor PJ34. Both PJ34 and CAF-1 ⁄ p60 siRNA exerted a striking effect in each cell line, as compared with untreated cells, that was most marked in CAL33 cells. This result suggests that the cooperative effect between PARP-1 and CAF-1 ⁄ p60 inhibitors is dampened in malignancy. In the near future, the association of ionizing radiation with radioenhancing anti-PARP-1 and ⁄ or anti-CAF-1 ⁄ p60 PARP-1, CAF-1 p60 and stem cells in OSCC 1101 agents may provide an opportunity to bypass the radioresistance of OSCC mCSCs, minimizing side effects in surrounding normal tissues. 69,91 OSCC immunohistochemical screening for PARP-1 high ⁄ CAF-1 p60 high ⁄ nestin high tumours could identify the highly radioresistant ⁄ chemoresistant cancers that would benefit from new molecular therapies, allowing a reduction in the severe morbidity and poor long-term survival of OSCC patients. 92 It is clear that there is still much to learn about OSCC biology. The occurrence of subsets of phenotypically distinct CSCs in primary tumours has been correlated with their aggressive behaviour. 13 However, the ultimate role of heterogeneity within the cancer cell population in determining tumour biology and the response to radiotherapy and chemotherapy remains to be fully clarified, before definitive decisions about the best stratification factors for personalized algorithms of OSCC treatment can be made. 93 Improved refining of the CSC population phenotype is necessary. We are aware that the arbitrariness of the cut-off scores for determining the positivity of potential tumour markers can hamper their prognostic value. So far, standardized scoring systems for evaluating immunohistochemistry in OSCC are lacking. It has been demonstrated that evaluation of immunoreactivity by using the percentage of positive tumour cells may be considered a reproducible scoring method with strong interobserver agreement. We preferred, instead, to use an unbiased method such as the median, to avoid potential bias through the use of less appropriate thresholds. Observer variation reached the acceptable standard for accurate assessment of protein expression. 94 To the best of our knowledge, this is the first report of simultaneous overexpression of CAF-1 p60, PARP-1 and stem cell markers in a malignant tumour. Our data indicate that the interaction between PARP-1, CAF-1 ⁄ p60 and nestin may constitute a specific hallmark of the aggressiveness of OSCCs. The high concordance between the expression levels of all of the evaluated proteins in whole sections and in TMAs may allow the rapid extension of the immunohistochemical evaluation of these markers to larger series of cases, favouring the adoption of screening for these proteins in the clinical setting. This may have important consequences for OSCC patients, providing us with novel candidate drug targets for reducing the number of deaths caused by the ineffective treatment of metastatic disease.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Table S1. Correlation between sex, age, localization, grading and staging and an adverse event (recurrence, metastasis, death from disease). Table S2. Correlation between expression of all markers and relapse. Table S3. Correlation between expression of all markers and metastasis. Table S4. Correlation between expression of all markers and death from disease.
Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.