Ku70 predicts response and primary tumor recurrence after therapy in locally advanced head and neck cancer

Abstract

5-Fluorouracil and cisplatin-based induction chemotherapy (IC) is commonly used to treat locally advanced head and neck squamous cell carcinoma (HNSCC). The role of nonhomologous end joining (NHEJ) genes (Ku70, Ku80 and DNA-PKcs) in double-strand break (DSB) repair, genomic instability and apoptosis suggest a possible impact on tumor response to radiotherapy, 5-fluorouracil or cisplatin, as these agents are direct or indirect inductors of DSBs. We evaluated the relationship between Ku80, Ku70 or DNA PKcs mRNA expression in pretreatment tumor biopsies, and tumor response to IC or local recurrence, in 50 patients with HNSCC. Additionally, in an independent cohort of 75 patients with HNSCC, we evaluated the relationship between tumor Ku70 protein expression and the same clinical outcomes or patient survival. Tumors in the responder group had significantly higher mRNA levels for Ku70, Ku80 and DNA-PKcs than those in the nonresponder group. Ku70 mRNA was the marker most significantly associated with response to IC. Moreover, high tumor Ku70 mRNA expression was associated with significantly longer local recurrence-free survival (LRFS). Ku70 protein expression was also significantly related to response, and patients with higher percentage of tumor cells expressing Ku70 had longer LRFS. In addition, the percentage of Ku70 positive cells, tumor localization and node involvement were significantly associated with overall survival of patient. Therefore, Ku70 expression is a candidate predictive marker that could distinguish patients who are likely to benefit from chemoradiotherapy or radiotherapy after the induction chemotherapy treatment, suggesting a contribution of the NHEJ system in HNSCC clinical outcome. © 2008 Wiley-Liss, Inc.

Therapy for locally advanced head and neck squamous cell carcinoma (HNSCC) aims at organ preservation, having displaced radical surgery. Concomitant platinum-based chemoradiotherapy (CRT) has become the standard treatment for this disease.1 5-Fluorouracil (5-FU) and cisplatin-based induction chemotherapy (IC), followed by CRT or radiotherapy (RT), has also been used to treat locally advanced HNSCCs, after demonstrating a benefit for organ preservation, loco-regional control and overall survival.2, 3

Tumor response to IC predicts response to RT and patient survival in locally advanced HNSCC.4–6 The IC response distinguishes sensitive carcinomas, which will follow a conservative treatment, from nonsensitive carcinomas, which will be treated with surgery. Conservative treatment consists of the administration of CRT (formerly RT alone) aimed at preventing the mutilating effect of surgery.

Despite conventional TNM information having a strong prognostic value in HNSCC,7 few predictive molecular markers of response to therapy exist in this pathology. Consequently, uncovering predictors of response to IC may improve our ability to anticipate tumor response to subsequent CRT, identifying patients who could benefit from a conservative treatment.

The cytotoxicity of the classic antineoplasic agents (5-FU, cisplatin, radiotherapy) depends of their ability to produce DNA damage in tumor cells. Despite these agents producing different types of DNA damage, all of them directly or indirectly produce DNA double strand breaks (DSB).8–10 CDDP produces interstrand crosslinks, stops DNA replication and generates DSBs.11 5-FU is an analog of uracil that is converted intracellularly to fluorodeoxyuridine monophosphate (FdUMP), a potent inhibitor of the enzyme thymidylate synthetase. This results in a deoxynucleotide (dNTP) pool imbalance, which disrupts DNA replication, generating DSBs.10 DNA DSB is the most toxic and mutagenic of all DNA lesions. Two mechanisms exist for DSB repair: homologous recombination and nonhomologous end joining (NHEJ). NHEJ has been described as the predominant DSB repair mechanism in mammalian cells.12

We hypothesized that the expression of NHEJ DNA repair (Ku70, Ku80 and DNA-PKcs) genes will influence tumor response to therapy on the basis of the following: (i) A high frequency of chromosomal translocations and alteration of double strand break (DSB) DNA repair in HNSCCs13–15; (ii) Direct induction of DSBs by radiation9 or indirectly by 5-FU10, 16 or cisplatin8, 11; (iii) The major importance of NHEJ in DSB repair, genomic stability maintenance and suppression of translocations in mammalian cells17, 18; (iv) An increased in vitro sensitivity to radiation by NHEJ protein inactivation and recovery of resistance by restoration of activity19–21 and (v) Involvement of Ku70, Ku80 and DNA-PKcs in apoptotic signaling after radiation- or chemotherapy-induced DSBs.22–24

Patients with locally advanced (Stage III or IV) HNSCC, treated at our institution with induction chemotherapy followed by either CRT/RT or surgery depending on tumor response, participated in a prospective mRNA study (n = 50) or in an independent retrospective immunohistochemistry (IHC) study (n = 75). We evaluated the relationship between Ku70, Ku80 or DNA-PKcs mRNA, or Ku70 protein expression in pretreatment tumor biopsies, and the degree of tumor response to IC, local recurrence and patient survival. Our aim was to test the capacity of the expression of these genes in identifying patients likely to benefit from a conservative treatment.

5-FU, 5-fluorouracil; CDDP, cis-diamminedichloridoplatinum; CRT, chemoradiotherapy; CT, computed tomography; dNTP, deoxynucleotide; DSB, double strand break; ES, embryonic stem; FdUMP, fluorooxyuridinine monophosphate; HNSCC, head and neck squamous cell carcinoma; HPV, human papillomavirus; HR, hazard ratio; HSCSP, Hospital de la Santa Creu i Sant Pau; IC, induction chemotherapy; IHC, immunohistochemistry; LRFS, local recurrence-free survival; MEFs, mouse embryo fibroblasts; MR, magnetic resonance; NHEJ, nonhomologous end joining; OS, overall survival; ROC, receiver-operating-characteristics; RT, radiotherapy.

Patients and methods

Patient characteristics and treatment plan

Accrual for the mRNA prospective study (n = 50) was initiated in 2002. Patients in the IHC retrospective study (n = 75) were treated during the 1995–2003 period at the Hospital de la Santa Creu i Sant Pau (HSCSP). All patients had pathologically confirmed, untreated, locally advanced (Stage III or IV) HNSCCs. The HSCSP Ethics Committee approved the study.

In both studies, patients were treated with IC consisting of the administration of cisplatin at a dose of 100 mg/m² on day 1, and 5-FU at a dose of 1,000 mg/m²/day by continuous intravenous infusion on days 2–6 every 3 weeks, per 3 courses. Patients who presented a high tumor response after IC followed a conservative treatment, which consisted of the administration of RT or CRT. Patients treated from 1995 to 2002 who presented a high tumor response after IC followed treatment with radiotherapy (RT). CRT was introduced in 2003 as an alternative to RT to treat patients with a substantial IC response. Progressively, CRT displaced RT as the treatment of choice after IC.

RT was administered to the primary tumor and clinically positive nodes in 35 fractions of 2 Gy, each over a 7-week period at a total dose of 70 Gy. Nodal areas not clinically involved by tumor received a total dose of 50 Gy. CRT consisted of RT at the same doses plus 3 cycles of cisplatin at a dose of 100 mg/m² on day 1 every 3 weeks.

Patients without significant response to IC were usually treated with surgery followed by RT. The characteristics of patients included in both studies are summarized in Table I.

Table I. Characteristics of Patients Included in the Prospective and Retrospective Studies

Patients characteristics	Prospective mRNA study (n = 50)	Retrospective IHC study (n = 75)
Sex
Men	45 (90%)	71 (95%)
Women	5 (10%)	4 (5%)
Tumor localization
Oral cavity	11 (22%)	7 (9.3%)
Oropharynx	7 (14%)	18 (24%)
Hypopharynx	14 (28%)	12 (16%)
Larynx	18 (36%)	38 (51%)
Lynph node involvement
Present	40 (80%)	41 (55%)
Absent	10 (20%)	34 (45%)
Tumor size (T)
2	8 (16%)	9 (12%)
3	24 (48%)	50 (67%)
4	18 (36%)	16 (21%)
Histological tumor grade
Good	5 (10%)	2 (3%)
Moderate	39 (78%)	65 (87%)
Poor	6 (12%)	8 (11%)
Tumor staging (TNM)
III	12 (24%)	40 (53%)
IV	38 (76%)	35 (47%)
Treatment subsequent to IC
Radiotherapy	16 (32%)	38 (51%)
Chemoradiotherapy	15 (30%)	4 (5%)
Surgery	19 (38%)	33 (44%)

RNA extraction, cDNA synthesis and quantitative PCR

The prospective study was performed using fresh pretreatment primary tumor biopsies obtained from patients with HNSCC. A sample aliquot was used for pathological diagnosis of the malignancy. Samples with a low content of tumor tissue were excluded from the study. Another aliquot was frozen immediately after taking the biopsy, in cold isopentane (histobath), and placed in liquid nitrogen until RNA processing.

Total RNA was extracted with Trizol® (Invitrogen, Paisley, UK) and the phenol–chloroform method. Samples were then precipitated twice with isopropanol and 70% ethanol, and cleaned-up using RNeasy® Spin columns (Qiagen, Valencia, CA). Total RNA was quantified espectrophotometrically. Reverse transcription was performed using 1.5 μg of total RNA and the High Capacity cDNA Archive Kit (Applied Biosystems, Foster City, CA), in a 50 μL final reaction volume, containing 5 μL of 10× RT buffer, 2 μL of 25× dNTPs mixture, 5 μL of 10× Random Hexamer Primers, 125 U of Multiscribe Reverse transcriptase and 40 U of RNase inhibitor (Invitrogen, Paisley, UK). These mixtures were incubated at 25°C for 20 min, and then at 37°C for 2 hr. Finally, heating at 95°C for 3 min inactivated the reverse transcriptase.

mRNA expression was measured on an ABIPrism 7000 Sequence Detection System (Applied Biosystems, Foster City, CA), using predesigned Taqman® Gene Expression Primer and probe Assays (Applied Biosystems, Foster City, CA), which were available for Ku70, Ku80, and β-actin (http://www.appliedbiosystems.com). All probes were FAM-labeled, except for β-actin that was VIC-labeled. For DNA-PKcs detection, we used Primer Express software v2.0 (Applied Biosystems, Foster City, CA) to design the forward (5′-TGGGAGCATCACTTGCCTTTAATAA-3′) and reverse (5′-CAAACTGTTCCACCAGAGACTCTT-3′) primers and a Taqman® probe (5′-CTTCCCTGAATTCCC-3′) (Table II).

Table II. Description of the Amplification and Hybridization Regions for mRNA Gene Expression Analysis by qPCR in the Prospective Study

Gene	Assay1	RefSeq2	Interexonic union	Amlicon size (bp)
1 Number assigned to predesigned Taqman Gene Expression Assay (Applied Biosystems; http://www.appliedbiosystems.com). 2 mRNA Reference Sequence amplified using Taqman Gene Expression Assays. Available at the NCBI reference Sequence Database (Refseq). 3 Custom-made sequence (see Patients and Methods section).
Ku70	Hs00750856_s1	NM_001469	Exon 1	99
Ku80	Hs00221707_m1	NM_021141	Exon 2/Exon 3	72
DNA-PKcs	–3	NM_006904	Exon 27/Exon 28	72
β-actin	Hs99999903_m1	NM_00101	Exon 1	171

mRNA quantitation was performed, in duplicates, in 20 μL total volume PCR reactions, using 2 μL of each sample cDNA, 10 μL of 2× Universal Taqman Master Mix, 1 μL of primers and probe mixture and 7 μL of H₂O, following the manufacturer's recommendations (Applied Biosystems, Foster City, CA). Thermal cycling conditions were 2 min at 50°C, 10 min at 95°C, and 40 cycles at 95°C for 15 s and 60°C for 1 min.

Gene expression results were calculated applying the comparative CT method (2^−(ΔΔCt)) as previously described.25, 26 Using the 2^−(ΔΔCt) method, the data are presented as the fold change in gene expression normalized to an endogenous gene and relative to a calibration sample. β-actin was used as the endogenous control to normalize the PCR results for the amount of RNA added to the reverse transcription reaction. We extracted RNA from the HNSCC UM-SCC-22A cell line (a gift from Dr. R.H. Brakenhoff), which was used as the calibration sample.27 First, we subtracted the endogenous gene expression from target gene expression (ΔCT = CT_{target gene(Ku80, Ku70 or DNA-PKcs)} − CT_{endogenous gene(β-actin)}) and, afterwards, calculated the expression in the tumor relative to the calibration sample (ΔΔCT = ΔCT_{tumor sample} − ΔCT_{calibrator(UM-SCC-22A)}). Each gene measure was expressed in relation to the level of the same gene in the UM-SCC-22A cell line. This cell line was maintained in 10% FBS DMEM (Invitrogen, Paisley, UK), with 2 mmol/L glutamine, 50U/mL penicillin and 50U/mL streptomycin, at 37°C in a humidified atmosphere containing 5% CO₂.

Immunohistochemistry

We used pretreatment formalin-fixed and paraffin-embedded biopsies of primary HNSCCs, with high tumor tissue content, to perform the retrospective study. Four-micron sections were deparaffinized in xylol and rehydrated using decreasing ethanol concentrations (100, 96, 80, 70 and 50%) and distilled H₂O. The samples were immersed in 1× Target Retrieval Solution pH = 6 (DakoCytomation S.A., St Just Desvern, Spain) and autoclaved over 10 min at 121°C for antigenic retrieval. Endogenous tissue peroxidase was inactivated by incubating samples in a 3% H₂O₂ solution for 10 min. Samples were then incubated with a mouse monoclonal antibody against Ku70 (Ab-4) (Lab Vision, Freemont, CA) at a 1:200 dilution. For primary antibody detection, we used the EnVison + Dual Link System-HRP Kit in an automatic Autostainer System (DakoCytomation S.A., St Just Desvern, Spain), according to manufacturer's instructions. Counterstaining was performed with hematoxylin (DakoCytomation S.A., St Just Desvern, Spain). Samples were dehydrated in a growing ethanol and xylol gradient, and mounted with DPX media (Sigma Aldrich, Tres Cantos, Spain). Negative controls were processed substituting the primary antibody by nonimmunized mouse serum. A normal mucousa and 2 HNSCC samples were used to control for IHC batch staining variability.

Immunohistochemistry analysis

We took three 100×-magnified images per sample, using a DP50 camera and an Olympus DX51 microscope, under the same lighting and time exposure conditions. The ViewFinder Lite v1.0 and Studio Lite v1.0 software (Olympus, USA) were used to capture and store all images. Afterwards, we eliminated all areas containing no tumor cells, using AdobePhotoshop software v7.0 (AdobePhotoshop systems, USA). We quantitated the IHC staining of the areas containing tumor cells, using the Metamorph v 5.0 (Universal Imaging, Downingtown, PA) software and applied the HSI (HUE-saturation-intensity) model,28, 29 which selects a particular area according to its color, as defined by its HUE (the dominant wavelength of transmitted light), saturation and intensity. For each pixel in an image, the values of HUE, saturation and intensity are independently transformed to one of 256 integral values within a 0–255 range. Setting a range between 0 and 255 for each of these 3 parameters makes it possible the selection of the particular areas of interest.28, 29

Evaluation of tumor response, local recurrence and patient survival

In both, the prospective and retrospective studies, tumor response was evaluated by comparing tumor volume before IC and after the third cycle of IC. Response was defined as a reduction in primary tumor volume, as measured by physical examination, fiber optic laryngoscopy, and CT scan/MR imaging following RECIST criteria.30

The responder group was composed of tumors presenting a complete response or a partial response higher than 50%. The nonresponder group included all tumors presenting a stable disease (<25% progression or <50% reduction) or progressive disease (>25% of progression in tumor volume).

We recorded local recurrence-free survival (LRFS), which was defined as time from diagnosis to recurrence at the primary location of the tumor. In addition, we recorded overall survival, defined as the time from diagnosis to patient death. The median follow-up time in the prospective study was 2.1 years. The median follow-up time for the retrospective study was 4.0 years. The molecular analysis was performed with no knowledge of clinical outcomes.

Statistical analysis

We compared tumor mRNA levels and the percentage of positively stained tumor cells between the responder and nonresponder groups by applying the Mann–Whitney U test. We performed a nonparametric receiver-operating characteristics (ROC) analysis to evaluate the diagnostic usefulness of Ku80, Ku70 and DNA-PKcs mRNA levels and Ku70 protein levels to discriminate between responding and nonresponding tumors. We established a cut-off level for mRNA and protein levels for each studied variable to distinguish tumors with high or low expression levels. These cut-off values were determined selecting the most accurate values obtained from the nonparametric ROC analysis, taking into account the best balances between sensitivity and specificity. Logistic regression analysis was used to evaluate the associations of Ku70, Ku80 and DNA-PKcs expression above or below the defined cut-off values, tumor size (T), node involvement and tumor localization, with IC response. Adjusted LRFS curves were estimated using the Kaplan–Meier method. We applied a 2-tailed log-rank test to evaluate the differences in LRFS between patients with tumor expression above or below the defined cut-off values. The association of Ku70, Ku80 and DNA-PKcs gene expression, node involvement, tumor size and localization with LRFS and OS was assessed applying a univariate and a multivariate Cox regression model analysis.

Differences were considered significant at p values <0.05 in all applied statistical tests. All statistical analyses were performed using the SPSS software v.14.01 (SPSS, Chicago, IL).

Results

Differences in Ku70, Ku80 and DNA-PKcs mRNA levels between responder and nonresponder tumors (prospective study)

The characteristics of the patients included in the prospective study are summarized in Table I. Twenty-eight out of 50 patients had a tumor response to IC higher than 50% (responder group), whereas 22 tumors showed responses lower than 50% (nonresponder group). The median mRNA level was 4.4 (range 0.5–309.8) for Ku70, 2.3 (0.1–36.8) for Ku80 and 3.3 (0.1–59.9) for DNA-PKcs when considering all samples (responders and nonresponders) included in the prospective study. The median mRNA level in the responder group was 6.6 (range 1.2–236.4) for Ku70, 3.2 (range 0.7–36.8) for Ku80 and 4.2 (range 0.2–59.9) for DNA-PKcs. The median mRNA level in the nonresponder group was 3.3 (range 0.5–309.0) for Ku70, 1.8 (range 0.1–3.3) for Ku80 and 2.4 (range 0.1–11.5) for DNA-PKcs.

We observed significantly higher mRNA tumor values for Ku70 (p = 0.005), Ku80 (p = 0.002) or DNA-PKcs (p = 0.017) in responders than in nonresponders (Figs. 1a–1c). Afterwards, we performed a ROC analysis to test the sensitivity and specificity of Ku70, Ku80 or DNA-PKcs mRNA levels in evaluating response to IC (Fig. 1d). The areas under the curve (AUC) were 0.73 [CI (95%) = 0.59–0.88, p = 0.005] for Ku70, 0.76 [CI (95%) = 0.63–0.90, p = 0.002] for Ku80 and 0.70 [CI (95%) = 0.55–0.88, p = 0.017] for DNA-PKcs. We next established a cut-off level between high and low mRNA levels for each gene by selecting the most accurate value obtained in the ROC analysis. Thus, we distributed all patients between 2 groups, one above and another below the following mRNA thresholds: 3.6 for Ku70, 2.6 for Ku80 and 5.0 for DNA-PKcs. The sensitivity, specificity and accuracy in predicting IC response with the established cut-offs were as follows: Ku70 (sensitivity 86%, specificity 68%, accuracy 78%), Ku80 (sensitivity 57%, specificity 77%, accuracy 66%) and DNA-PKcs (sensitivity 43%, specificity 86%, accuracy 62%).

Figure 1. Significant differences in Ku70 (a), Ku80 (b) and DNA-PKcs (c) mRNA levels in pretreatment biopsies between tumors with a response to induction chemotherapy higher than 50% (Resp > 50%) and tumors with response lower than 50% (Resp < 50%) in the prospective study. Differences in NHEJ gene expression between the responder group and the nonresponder group were compared using the nonparametric Mann–Whitney U test. (d) Curves obtained applying receiver-operating-characteristics (ROC) analysis for Ku70, Ku80 and DNA-PKcs.

Logistic regression analysis was conducted to evaluate the associations of mRNA levels, tumor localization, lymph node involvement and tumor size (T) with IC response. Only high (above 3.6) Ku70 mRNA levels were significantly associated with response to IC higher than 50% [p = 0.03; odds ratio 5.9; CI (95%) = 1.28–29.6]. Therefore, ROC and logistic regression analysis indicated that Ku70 mRNA gene expression was the best marker to discriminate between responding and nonresponding tumors.

Ku70 and Ku80 mRNA levels are associated with local recurrence-free survival (prospective study)

We next analyzed whether the mRNA expression of these genes was associated with primary tumor recurrence (Fig. 2a). Patients whose tumors expressed Ku70 mRNA levels above the established 3.6 threshold had a significantly higher probability of surviving without having primary tumor recurrence (increased LRFS) than patients bearing tumors with lower mRNA levels (p = 0.043). Similarly, patients whose tumors expressed Ku80 mRNA levels above the 2.6 threshold had a higher LRFS than patients bearing tumors with lower mRNA levels (p = 0.04). DNA-PKcs mRNA levels did not show an association with patient's LRFS (p = 0.457).

Figure 2. Analysis of patient LRFS in the prospective study. Differences in adjusted local recurrence-free survival (LRFS) between patients bearing tumors expressing pretreatment Ku70, Ku80 or DNA-PKcs mRNA levels above or below the depicted cut-off levels. (a) Analysis performed including all patients treated with induction chemotherapy, followed by either surgery or conservative treatment. (b) Analysis performed only with the group of patients who followed a conservative treatment after IC.

We also applied a Cox model analysis to study the association of mRNA levels, tumor size (T), localization, lymph node involvement and IC response, with LRFS (Table III). On a Cox univariate analysis, we found that Ku70 mRNA levels (p = 0.05) and Ku80 mRNA levels (p = 0.05) were associated with LRFS. On a multivariate Cox analysis, Ku70 mRNA levels (p = 0.04) and tumor size (p = 0.03) were significant independent risk factors of LRFS. Therefore, in our study, Ku70 gene expression remains the only marker able to consistently predict local control of the disease. Thus, patients with tumors showing high Ku70 mRNA levels presented a higher probability of surviving without primary tumor recurrence than patients with tumors showing low expression of this gene. Ku80 mRNA, despite displaying the same trend as Ku70 and being also a predictive marker of response, showed a weaker association with LRFS.

Table III. Hazard Ratios for Local Recurrence Obtained Applying Cox Model Analysis in the Prospective Study

	Local recurrence-free survival
	Univariate		Multivariate
	HR (95% CI)	p value	HR (95% CI)	p value
The bold values indicate the existence of statistically significant differences.
All patients
KU70 mRNA (<3.6 vs. >3.6)	2.8 (1.0–7.7)	0.05	4.7 (1.1–19.8)	0.04
KU80 mRNA (<2.6 vs. >2.6)	3.5 (1.0–12.4)	0.05	2.1 (0.5–9.3)	0.34
DNA-PKcs (<5.0 vs. >5.0)	1.5 (0.5–4.9)	0.46	1.7 (0.2–12.6)	0.59
Tumor size (T) (T3–T4 vs. T1–T2)	2.3 (0.8–6.6)	0.14	7.4 (1.3–42.9)	0.03
Node involvement (N^+vs. N^0)	2.1 (0.5–9.3)	0.33	1.0 (0.2–6.0)	0.97
Localization	1.4 (0.4–4.3)	0.60	2.1 (0.5–8.6)	0.32
Conservative treatment
Ku70 mRNA (<3.6 vs. >3.6)	6.9 (1.9–26.5)	<0.01	28.2 (1.7–47.0)	0.02
Ku80 mRNA (<2.6 vs. >2.6)	2.9 (0.7–11.8)	0.13	0.9 (0.0–8.6)	0.95
DNA-PKcs (<5.0 vs. >5.0)	1.7 (0.4–6.9)	0.45	0.4 (0.02–7.3)	0.53
Tumor size(T) (T3–T4 vs. T1–T2)	1.8 (0.4–7.2)	0.42	4.9 (0.7–36.1)	0.12
Node involvement (N^+vs. N^0)	2.6 (0.3–20.6)	0.37	1.3 (0.1–16.5)	0.85
Localization	1.1 (0.2–5.1)	0.94	0.5 (0.1–3.2)	0.45

After applying a Cox model analysis to assess the association of mRNA levels, tumor size (T), localization, lymph node involvement and IC response with overall survival, we observed that high Ku70 levels displayed the same trend towards associating with longer overall survival, as it happened with LRFS; however, the observed differences did not reach statistical significance (p = 0.14).

Ku70 mRNA levels are associated with local recurrence-free survival in patients who followed a conservative treatment (prospective study)

We next searched for the possible differences in adjusted LRFS within the subset of patients who received CRT or RT after IC (n = 31), excluding those treated with surgery after IC (n = 19) (Fig. 2b). Our objective was to study Ku70, Ku80 and DNA-PKcs as possible markers of tumor response and primary tumor recurrence after genotoxic therapy. Surgery is applied to patients with poor response to IC, and their inclusion could have altered the LRFS registered in patients under genotoxic treatment. All patients (n = 28) of the responder group followed a treatment with CRT or RT after IC. Nineteen patients of the nonresponder group underwent surgical excision of their tumors after IC. Three patients of the nonresponder group refused mutilating surgery and, contrarily to medical advice, were treated with RT or CRT. Out of the 31 IC patients who followed a conservative treatment (CRT or RT) after IC, those whose tumors expressed Ku70 mRNA levels above the 3.6 threshold had a significantly higher probability of surviving without having a primary tumor recurrence (increases LRFS) than patients bearing tumors with lower mRNA levels (p = 0.001). Moreover, patients whose tumors expressed Ku80 mRNA levels above the 2.6 threshold showed a trend towards increased LRFS, but it did not reach statistical significance (p = 0.113). DNA-PKcs mRNA levels did not show significant differences in patient's LRFS (p = 0.442).

A univariate Cox model analysis confirmed the significant association between Ku70 mRNA levels and LRFS (p < 0.01) (Table III). Moreover, a multivariate Cox model analysis showed that Ku70 mRNA was the only independent risk factor for local recurrence-free survival in patients who followed conservative treatment after IC (p = 0.02) (Table III).

On a multivariate analysis, the difference in relative risk of primary tumor recurrence between high and low Ku70 mRNA tumor patients was 6 times higher when only conservatively treated patients were included in the analysis than when all patients were included (HR: 28.2 vs. 4.7, Table III). A univariate (p = 0.23) and a multivariate (p = 0.64) Cox model analysis in patients treated with surgery after IC showed no association between Ku70 mRNA levels and LRFS. The capacity of Ku70 mRNA in predicting local recurrence after conservative treatment was improved when the data on LRFS after mutilating surgery were excluded from the analysis, indicating that Ku70 mRNA is a better marker of local recurrence after CRT/RT than of recurrence after surgery.

Differences in Ku70 protein expression between responder and nonresponder tumors (retrospective study)

Patient characteristics in the retrospective study are summarized in Table I. Out of 75 analyzed patients, 39 had a tumor response to IC greater than 50% (responder group), whereas 36 patients showed a tumor response lower than 50% (nonresponder group). Ku70 immunohistochemical analysis showed an exclusively nuclear staining pattern. Using the “Set color threshold” tool of Metamorph v5.0 software and the HSI color model (Fig. 3a), we established a 180–255 HUE range to select all brown primary antibody-stained areas (positive nuclei) (Fig. 3c), which ensured the absence of any selected area in negative control samples. Moreover, we established a 160–255 HUE range to select all (brown + blue) tumor cell nuclei present in the sample (Fig. 3d), leaving out cell cytoplasm, membrane or keratin deposits. Since all tumor nuclei present in each sample image displayed a similar size, we calculated the percentage of positive tumor cells dividing the area occupied by positive tumor nuclei by the area occupied by all nuclei. Two experienced pathologists validated this method.

Figure 3. Ku70 Immunohistochemistry quantification used in the retrospective study. (a) We used the “Set Color Threshold” tool, implemented in Metamorph software v5.0, to establish a suitable HUE range that selects areas presenting specific immunostaining characteristics. (b) Immunohistochemistry quantification was performed in 100× magnified images. Ku70 shows a nuclear staining pattern. (c) All positive immunostained nuclei, present in the image, were selected using a 180–255 HUE range (green areas). (d) All the nuclei (positive and negative), present in the image, were selected using a 165–255 HUE range (green areas).

Figure 4a shows Ku70 staining in 2 tumors from patients with different responses to IC treatment. The median of the percentage of Ku70 positively stained cells was 63.31 (range 7.4–92.8) for all samples (responders and nonresponders) included in the retrospective study. The median of the percentage of Ku70 positively stained cells in the responder group was 70.7% (range 19.0–92.8%) and in the nonresponder group was 57.7% (range 7.4–88.3%). We observed a significantly higher percentage of Ku70 positive tumor cells in responders than in nonresponders (p = 0.036) (Fig. 4b). Using ROC analysis, we obtained an AUC of 0.64 [CI (95%) = 0.51–0.77, p = 0.036] for Ku70 positive cells (Fig. 4c). The cut-off value between high and low percentage of positive Ku70 cells was established at 74%, which was the most accurate value obtained in the ROC analysis. Its sensitivity, specificity and accuracy in predicting IC response was 47, 69 and 59%, respectively. Thus, in pretreatment tumor samples, the percentage of Ku70 positive tumor cells was significantly associated with response to IC.

Figure 4. Analysis of Ku70 protein expression in the retrospective study. (a) Differences in Ku70 immunostaining in pretreatment biopsies between a representative tumor in the responder group and a representative tumor in the nonresponder group. (b) Significant differences in the percentage of Ku70 positive tumor cells between tumors with a response to induction chemotherapy higher than 50% (Resp >50%) and tumors with response lower than 50% (Resp <50%). (c) Receiver-operating-characteristics (ROC) curves applied to the percentage of Ku70 positive cells. (d, e) Differences in adjusted local recurrence-free survival (LRFS) between patients bearing tumors with a percentage of Ku70 protein staining higher or lower than 74%. (d) Analysis performed including all patients treated with induction chemotherapy, followed by either surgery or conservative treatment. (e) Analysis performed only in the group of patients who followed a conservative treatment after IC.

Association between the percentage of Ku70 positive cells and local recurrence-free survival (retrospective study)

Patients whose tumors contained a percentage of Ku70 positive cells above the 74% threshold had a significantly increased probability of surviving without having a primary tumor recurrence (longer LRFS), as compared to patients whose percentage of positive tumor cells were below this threshold (p = 0.013) (Fig. 4d). Cox univariate (p = 0.03) and multivariate (p = 0.02) analyses revealed the percentage of Ku70 positive tumor cells as the most significant factor associated with LRFS (Table IV).

Table IV. Hazard Ratios for Local Recurrence and Cancer Death Obtained Applying Cox Model Analysis in the Retrospective Study

	Local recurrence-free survival				Overall survival
	Univariate		Multivariate		Univariate		Multivariate
	HR (95% CI)	p value	HR (95% CI)	p value	HR (95% CI)	p value	HR (95% CI)	p value
The bold values indicate the existence of statistically significant differences.
All patients
%Ku70 positive cells (<74% vs. >74%)	5.2 (1.2–22.9)	0.03	5.6 (1.3–24.3)	0.02	3.9 (1.2–13.3)	0.03	5.1 (1.5–17.4)	<0.01
Tumor size (T) (T3–T4 vs. T1–T2)	2.7 (0.4–20.1)	0.34	2.7 (0.4–21.4)	0.34	1.4 (0.3–6.0)	0.65	2.2 (0.5–9.6)	0.30
Node involvement (N^+vs. N^0)	1.1 (0.5–2.8)	0.78	1.0 (0.4–3.0)	0.97	3.1 (1.1–8.3)	0.03	1.9 (0.6–5.8)	0.26
Localization	1.9 (0.8–4.9)	0.17	2.3 (0.8–6.7)	0.12	5.9 (2.0–17.4)	<0.01	5.6 (1.7–18.4)	<0.01
Conservative treatment
%Ku70 positive cells (<74% vs. >74%)	9.5 (1.3–73.1)	0.03	12.3 (1.6–96.6)	0.02	2.0 (0.7–13.9)	0.16	5.5 (1.1–26.5)	0.03
Tumor size (T) (T3–T4 vs. T1–T2)	3.3 (0.4–25.0)	0.26	4.6 (0.6–38.2)	0.16	2.4 (0.3–18.4)	0.42	5.6 (0.7–46.9)	0.11
Node involvement (N^+vs. N^0)	1.0 (0.4–2.9)	0.99	1.7 (0.5–5.2)	0.37	2.2 (0.6–8.1)	0.26	2.5 (0.6–10.2)	0.19
Localization	2.0 (0.7–6.0)	0.22	2.8 (0.9–9.0)	0.08	12.3 (1.6–96.6)	0.02	16.4 (2.0–134.3)	<0.01

Similar to the mRNA prospective study, in this study, we analyzed the adjusted LRFS within the subset of patients who received RT or CRT after IC, excluding patients who underwent a surgery procedure after IC. All patients (n = 39) of the responder group followed a treatment with CRT or RT after IC. Thirty-three patients of the nonresponder group followed a surgery treatment after IC. Three patients of the nonresponder group, contrarily to medical advice, rejected mutilating surgery and followed CRT or RT treatment. A total of forty-two patients followed a conservative treatment after IC. Patients whose tumors contained a percentage of Ku70 positive nuclei above the established 74% threshold had a significantly longer LRFS as compared to patients whose percentage of positive tumor nuclei were below this threshold (p = 0.008) (Fig. 4e). Cox univariate (p = 0.03) and multivariate (p = 0.02) analyses confirmed the association between the percentage of Ku70 positive tumor cells and LRFS in patients who followed conservative treatment after IC (Table IV). A univariate (p = 0.51) and a multivariate (p = 0.48) Cox model analyses in patients treated with surgery after IC showed no association between the percentage of Ku70 positive tumor cells and LRFS.

In summary, high percentage of Ku70 positive cells was significantly associated with a longer local primary-recurrence-free survival. In addition, the predictive capacity of Ku70 protein expression became more significant when analyzing only patients under conservative treatment than when including all patients (conservative plus surgical) in the analysis (see Figs. 4d and 4e).

Association between the percentage of Ku70 positive cells and overall survival (retrospective study)

On a Cox univariate analysis, we found that the percentage of tumor Ku70 positive cells (p = 0.03), node involvement (p = 0.03) and tumor localization (p < 0.01) were significantly associated with patient's overall survival (Table IV). Moreover, on a multivariate Cox analysis, only the percentage of Ku70 positive cells (p < 0.01) and tumor localization (p < 0.01) were significant independent risk factors of overall survival (Table IV). Performance of the same analysis only on patients who underwent conservative treatment after IC yielded similar results (Table IV).

Discussion

The aim of this work was to obtain predictive markers of response to therapy in locally advanced HNSCCs. In the prospective study, we have described that tumors showing a response higher than 50% to induction chemotherapy (IC) had significantly higher pretreatment Ku70, Ku80 or DNA-PKcs mRNA levels than tumors with a response lower than 50%. In the retrospective study, which was carried out in a distinct cohort of patients, we analyzed Ku70 tumor expression, because it was the best marker of response and local recurrence in the prospective analysis. This study included a higher number of patients and longer follow-up times allowing to perform a more extensive local recurrence and overall survival analysis. Here, our goal was to know if the observed relationship between Ku70 expression and the clinical variables described in the prospective study was also found in the retrospective study and whether these findings went in the same direction, despite using a different methodology (IHC instead of qPCR).

In the retrospective study, Ku70 was the most significant independent factor in predicting LRFS, being also associated with OS. In contrast, node involvement and tumor localization were associated with overall survival, but they did not predict LRFS. The observed differences between the LRFS and OS analyses suggest that clinical factors other than primary tumor recurrence are involved in determining overall survival. Indeed, node involvement and distant metastases, together with primary tumor recurrence, are the factors showing the highest association with poor prognosis in patients with HNSCC.7, 31

These results support the notion that Ku70 expression is a predictive marker of response and recurrence after adjuvant therapy. In this sense, it differs from classical prognostic factors such as tumor localization or node involvement, which associate with clinical outcomes independently of the applied treatment. Thus, prognostic markers are related to tumor aggressiveness and to the nature of the mutations driving cell growth, motility and dissemination capacity, which may not necessarily relate to tumor response to therapy. In contrast, predictive markers are related to the interaction between the tumor and the therapeutic agents and anticipate tumor response. In HNSCC, prognosis largely depends on conventional TNM information, which stratifies patients in terms of tumor aggressiveness, and indicates the requirement or not for systemic therapy. However, this information has so far proven inadequate in predicting response to nonsurgical therapy.7, 32 Thus, despite lymph node involvement is the single most adverse independent prognostic factor31; it is, however, not useful in predicting response to adjuvant therapy.32

Our finding of Ku70 gene or protein expression as a marker of tumor response to IC, as well as a marker of local disease control and patient survival after genotoxic therapy, suggest that this protein contributes to determine tumor response. These associations are consistent with the previous demonstration that tumor response to IC predicts response to subsequent RT4, 33 and patient survival.5 To our knowledge, no previous report has addressed the prediction of response to therapy by NHEJ genes/proteins in patient biopsies of locally advanced head and neck squamous cell carcinoma (HNSCC) treated with chemoradiotherapy. We are aware of only 2 related studies, 1 performed in primary cultures from mostly Stage IV head and neck carcinoma biopsies, which found no correlation between DNA-PKcs protein levels and in vitro radiosensitivity.34 Another study evaluated DNA-PKcs, Ku70 and Ku80 expression in head and neck cancer patients treated with radiotherapy and found no relationship with radiosensitivity.35 Nevertheless, there are some clinical reports regarding NHEJ protein prediction of response to therapy on related tumor types. Thus, in agreement with our results, high expression of DNA-PKcs, detected by immunohistochemistry, is associated with response to chemoradiation in esophageal carcinomas.36 Similarly, high expression of Ku80 or DNA-PKcs protein associates with increased survival in tonsillar carcinoma patients treated with radiotherapy.37 In contrast to these and to our results, high levels of Ku70 or DNA-PKcs, measured by IHC, associate with lower locoregional control after concurrent chemoradiotherapy in patients with nasopharyngeal carcinoma38; nevertheless, this is considered a tumor entity different from HNSCCs.39 In addition, results predicting response by Ku70 and/or Ku80 proteins by IHC, in a direction opposed to our findings, are also found in other tumor types such as cervical carcinoma, since high levels of these proteins associate with lower response and survival to radiotherapy in patients with cervical cancer.40, 41

Cell type-dependent response to therapy involving NHEJ proteins

Despite our identification of Ku70 and, to a lesser degree, Ku80 or DNA-PKcs as possible predictive markers of response to CRT in HNSCC, our findings went in a direction contrary to that anticipated. We were expecting higher NHEJ protein levels, or increased DNA-PK complex activity, being related to enhanced DNA damage repair, which in turn would lead to lower tumor responses to therapy. This assumption is based on previous in vitro and clinical reports involving DNA repair proteins. For instance, in vitro inactivation of NHEJ proteins associates with higher radiosensitivity, whereas restoration of repair activity restores resistance.21 Moreover, high Ercc1 levels predict for poor response in nonsmall cell lung42 or ovarian43 carcinoma patients.

Searching for a possible mechanistic explanation of the unanticipated direction of our findings, we have exhaustively reviewed the in vitro and in vivo literature that evaluates the role of NHEJ proteins in response to DNA damage. Despite recognizing inconsistency of our results with some previous work, we also found solid work agreeing with our findings. Overall, the literature reports that response to genotoxic therapy by DNA repair proteins depends on the studied cell type,44 on its differentiation state45 or even on its degree of functional activation.46 In the following, we are describing findings, in specific cell types, in which inactivation, or low level of activity, of the NHEJ system decreases their sensitivity to genotoxic agents, and its possible mechanistic basis. Afterwards, we are describing results in other cell types, in which inactivation of NHEJ proteins leads to increased cell sensitivity.

In agreement with the direction of our findings, inactivation of Ku70, in the chicken B lymphocyte cell line DT40, significantly increases their viability when exposed to high doses of double strand break (DSB) inducers, such as γ-radiation or methyl methanesulfonate, as compared with wild type cells.47 Similarly, enhanced cell survival, through suppression of p53-dependent apoptosis, has been reported in thymocytes of DNA-PKcs−/− mice treated with whole body-ionizing radiation, as compared to wild type mice.48 In addition, DNA damage-induced apoptosis by ionizing γ-radiation is abolished in DNA-PKcs−/− mouse embryo fibroblasts (MEFs) expressing E1A.49 Moreover, cisplatin induces marked cell death in Ku80+/+ immortalized MEFs, Ku80+/+ Chinese hamster ovary-derived (CHO) cells or SCID cells complemented with human DNA-PK, as compared with their matched deficient counterparts. This cisplatin induced-death is mediated by the kinase function of the DNA-PK complex and conveyed to neighboring cells through gap junctions, whereas cells deficient in Ku80 or DNA-PKcs are markedly resistant to the drug.24 In agreement with a role for the DNA-PK complex proteins in apoptosis, the exposure of MEFs or glioma cell lines to γ-radiation leads to DNA-PK and Chk2 phosphorylation of p53, which mediates subsequent induction of apoptosis.50, 51 Similarly, the apoptosis induced by IGFBP-3 in glioblastoma or prostate cancer cell lines is blocked when DNA-PK is knocked out or chemically inhibited.52 In addition, in breast cancer cells exposed to ionizing radiation, a nuclear trimeric protein complex, including clusterin, Ku70 and Ku80, is formed that constitutes a death signal for severely damaged cells.53

In contrast to the results described above, the role of the NHEJ proteins in DNA repair is consistent with the association between low Ku70 expression or DNA-PK activity and sensitization to radiation in fourteen esophageal cancer cell lines.54 Similarly, downregulation of Ku70 or DNA-PKcs by siRNA induces sensitization to cisplatin, etoposide or topotecan in the human cervical carcinoma HeLa cell line.20 Also, in low-passage normal human fibroblasts, siRNA knockdown of DNA-PKcs resulted in increased radiosensitivity.19 Moreover, enhanced NHEJ DNA repair activity is associated with lower sensitivity to genotoxic agents in primary cultures of human B-chronic lymphocytic leukemia cells, so that treatment with DNA-PK inhibitors increases their sensitivity to the DSB inducers γ-radiation, etoposide or neocarzinostatin.55 Similarly, Ku70−/− embryonic stem cells have showed an increased sensitivity to γ-radiation as compared to Ku70 heterozygous or wild type embryonic stem cells.56 These findings suggest that NHEJ proteins may function in DNA repair.46, 57 Moreover, DNA-PKcs may also participate in a NFkB-dependent antiapoptotic pathway that protects cells from death induced by toposisomerase inhibitors, in SV40-transformed human fibroblasts or CHO cells or p53 null MEFs.58 Similarly, DNA-PKcs inhibits apoptosis induced by heat shock treatment in HeLa, CHO or mouse lung fibroblast cells, a distinct function from its involvement in DNA repair.59

Therefore, the apparent contradiction between the results supporting and opposing our findings is solved when considering the existence of a cell type-dependent response to genotoxic therapy. Consistent with this notion, response to DNA damage by double strand break (DSB) inducers, involving the NHEJ system, is cell type-dependent.46 Thus, a cell type-dependent response to DNA damage, mechanistically involving the DNA-PK complex, has been proposed on the basis of 2 main and alternate functions. The activation of the DNA-PK complex by radiation or other genotoxic agents could trigger signaling pathways that result in apoptosis or cell cycle arrest.44 Cells, whose physiological function involves their rapid proliferation (e.g., lymphocytes), may require active apoptotic pathways to avoid tumorogenesis after genotoxic damage; in contrast, cells playing a supportive role to other cell types (e.g., stroma providing growth factors to epithelial cells) may instead enter senescence, to maintain their function while avoiding transformation.44 In agreement with the described cell type dependency, cells with the same genetic background, but at distinct differentiation states, respond differently to genotoxic agents, in the same mouse model. Thus, DNA-PKcs−/− embryonic stem (ES) cells show a similar level of sensitivity to IR than wild type ES, whereas DNA PK−/− fibroblasts (differentiated) cell line derived from the same mice are significantly more sensitive to IR than its wild type counterpart.45 On the other hand, some cell types change the NHEJ protein function, as they change their activation status. Thus, in nonactivated human multiple myeloma (MM) cells, Ku80 confers sensitivity to DNA damage; however, CD40 activation of MM cells induce Ku80 and Ku70 translocation to the plasma membrane changing their function towards antiapoptosis, which then leads to protection against apoptosis triggered by irradiation or doxorubicin.60

Further support for a cell type-dependent role of NHEJ in response to genotoxic therapy comes from findings in some cell types, in which the DNA complex does not appear to play any role in determining response to genotoxic agents. Thus, the sensitivity to IR in mouse ES cells knock out for Artemis, a member of the NHEJ pathway, does not vary as compared to ES wild type cells.61 Similarly, DNA-PK-deficient murine SCID embryogenic fibroblast or the DNA-PKcs−/− human glioma cell line MO59J shows a similar sensitivity to etoposide as wild type MEFs or the DNA-PKcs+/+ MO59K glioma cell line.62 Moreover, in human lymphoblastic cell lines, siRNA knockdown of DNA-PKcs results in no significant increase in radiosensitivity.19

We also reviewed whether other proteins involved in DNA repair followed a cell type-dependent response to genotoxic agents. We choose to focus on p53 role, because it is the most extensively studied gene regarding response to chemotherapy and/or radiotherapy, and because of its involvement in processes similar to those of the NHEJ proteins (DNA repair, apoptosis, genomic stability) and of its functional link with NHEJ signaling.48–50 We observed a pattern of response to DNA damage similar to that described above for the NHEJ system. Thus, p53 mutation may be a predictive marker of response to 5-FU and cisplatin-based neoadjuvant chemotherapy in HNSCC patients, since tumors with no response have a significantly higher prevalence of p53 mutations than responder tumors.63 Moreover, we also found a clear cell-type-dependent pattern of response to genotoxic agents in tumors other than HNSCC, as a function of their p53 status. Thus, inactivation of p53 shows either sensitivity or resistance to DNA damage depending on the tumor type, cancer cell line or primary cell line being tested.57 Tumors bearing a p53 wild type gene show higher response to radiotherapy/chemotherapy, which associates with longer patient survival, while its mutational inactivation leads to lower response and survival in nonsmall cell lung,64 breast65 or ovarian66 carcinoma. In contrast to these results, an increased response in p53 mutant testicular,67 glioma68 or bladder69 tumors has been observed, as compared to p53 wild type tumors.67 Moreover, and similar to NHEJ proteins, p53 functions in DNA repair as well as in apoptotic regulation,57 which could lead to contrary outcomes regarding response to genotoxic therapy. Therefore, some of the p53 results that have been reported could be unexpected if only the participation of p53 in DNA repair was considered rather than acknowledging a complex and cell-dependent role in DNA repair and apoptosis.

Overall, the literature reports that cell response to genotoxic therapy in human tumors, or in in vitro and in vivo models, after the inactivation of NHEJ proteins (and also of p53) depends on the studied cell type, and or its differentiation state or functional activation. These studies demonstrate the increasing complexity of this area of research; thus, in addition to its function in DNA repair and stress response, the proteins of the DNA-PK complex (DNA-PKcs, Ku70 and Ku80) are implicated in multiple and/or separated pathways that regulate distinct cell death pathways.

In summary, our results in HNSCCs could be explained considering that, in addition to their expected DNA repair function, the NHEJ proteins participate in cell apoptotic regulation. Thus, in a particular cell type (e.g., head and neck carcinoma), higher levels of each of these proteins could enhance apoptosis (signaling through the pathways available for this cell type, considering that they remain active after transformation) and lead to a higher tumor response. In contrast, in other cell types, in which apoptotic pathways are blocked, or in which NHEJ proteins participate only in repair functions, higher levels of these proteins would lead to increased repair and lower tumor responses. Therefore, the crosstalk between the available repair and apoptotic functions of the specific repair system (e.g., NHEJ) in a specific cell type (e.g., HNSCC) could determine if the treated cell is sensitive or resistant to the damaging agent (e.g., DSB inducer). Nevertheless, the complexity of this area requires the mechanistic dissection of the candidate pathways, as they relate to our results, before we could establish a definite role for some of the NHEJ proteins in HNSCC cells. We are now starting to address this issue.

Clinical implications of our findings

Independent of the exact role that Ku70 gene plays, our results support its use as a predictor of response to therapy, primary tumor recurrence and patient survival. Moreover, our results need to be validated in independent studies before their clinical introduction, as it should happen with other proposed markers (e.g., p53 or HPV infection), which are not still in use.63, 70, 71 Finally, in an attempt to extend their possible predictive capacity, we are now evaluating whether NHEJ genes may also predict HNSCC response to primary concomitant CRT.

In summary, our results suggest that Ku70 mRNA could be a good candidate to be validated as a predictive marker of response to genotoxic therapy. Thus, in biopsies of locally advanced HNSCCs, the levels of tumor mRNA or the percentage of protein positive tumor cells for Ku70 can identify patients with high probability of response to induction chemotherapy and longer local recurrence-free survival, leading to longer overall survival. These patients would benefit from a conservative treatment based on chemoradiotherapy or radiotherapy and would differ from those who would not respond and would require surgical resection or the exploration of alternative therapies.

Acknowledgements

We would like to thank Mr. Luis Carlos Navas for his technical support and Dr. Montserrat Lopez for her collaboration in biopsy sampling. We also wish to thank the patients who gave permission to use their tissue for research.