Predictive and prognostic significance of telomerase levels/telomere length in tissues and peripheral blood in head and neck squamous cell carcinoma

A growing body of evidence indicates that the expression of TERT, the catalytic subunit of telomerase, is a biological marker of progression in several cancers. We investigated the predictive and prognostic role of TERT levels and telomere length in tissues and peripheral blood in patients with head and neck squamous cell carcinoma (HNSCC). High TERT levels in cancer tissues were independently associated with worse response to therapy (odds ratio [OR]:6.26), regional failure (hazard ratio [HR]:5.75), progression (HR:2.12), and death (HR:3.53). Longer telomeres in the mucosa surrounding the tumor (SM) were independently associated with a lower risk of mucosal failure (HR:0.39). While telomere length in peripheral blood mononuclear cells (PBMC) significantly decreased with age, no correlation was found between age and telomere length in SM. No associations were found between TERT levels in plasma and telomere length in PBMC and the prognostic variables. High levels of TERT transcripts in cancer cells represent a reliable prognostic marker for identifying HNSCC patients with risk of progression. The altered relationship of telomere length to age in SM compared with PBMC suggests that in a subset of cases the phenotypically normal SM constitutes an acquired telomere-shortened epithelial field prone to genetic instability.


Results
Demographic and clinical characteristics of patients. The clinical characteristics of the 101 patients are summarized in Supplementary Table 1. The study group had a median age of 65 years (range, 28-87 years) at presentation and included 74 (73%) male and 27 (27%) female patients. The female prevalence was significantly lower among tumors developing from the hypopharynx/larynx (P = 0.003; Supplementary Table 1). The majority of patients were ever smokers (73%) and ever drinkers (59%). However, ever drinking was less frequent in patients with oropharyngeal SCC (P = 0.018; Supplementary Table 1), which overall were more commonly diagnosed in advanced stage (P = 0.003; Supplementary Table 1). Six percent (n = 4/66) of patients submitted to primary surgical treatment had positive/close margins. Among the 66 patients treated with upfront surgery, 28 (42%) received adjuvant (chemo)radiotherapy.
Telomere length in tumor, SM, and PBMC. Median telomere length in tumor cells, SM, and PBMC according to descriptive characteristics are reported in Table 1. The telomere length in tumor cells and in PBMC was not significantly associated with any of the descriptive variables investigated herein. On the other hand, telomere length in SM was significantly shorter in male patients (P = 0.032) and in ever drinkers (P = 0.004). No correlation was found between age and telomere length (Table 1). However, a finer classification of age revealed that telomere length in PBMC, but not in tumor and SM (Fig. 1A,B), significantly decreased with age (P = 0.037; Fig. 1C); no correlation was found between telomere length in SM and PBMC (r-Spearman = 0.22; P = 0.107). Additionally, no correlation existed between telomere length in SM and in the tumor, nor between telomere length in PBMC and in the tumor.
TERT levels in tumor, SM, and plasma. Median levels of TERT in tumor, SM, and plasma according to descriptive characteristics are reported in Table 1. TERT levels in tumor cells were significantly higher in females (P = 0.023) and in patients with advanced stage disease (P < 0.001) and neck metastases (P < 0.001). TERT levels in SM were significantly higher in patients with cancer of the hypopharynx/larynx (P = 0.013) and in those with advanced stage disease (P = 0.002). Among oropharyngeal cancers, patients with HPV-driven oropharyngeal cancer reported higher TERT levels in SM (P < 0.001). Circulating TERT levels in plasma were estimated in 97 cases; 42 cases, including 18 of those who had high tumor TERT levels, were negative for TERT mRNA. Overall, the analysis of TERT levels in plasma did not yield significant results (Table 1) Short telomeres in SM were associated with worse mucosal control (5-year mucosal control = 62% vs 81% in patients with telomere length in SM < 1.0675 vs > 1.0675; P = 0.027; Fig. 2A). This association was confirmed by multivariate analysis (Table 2) after adjustment for clinical variables, with a HR for mucosal failure of 2.57 (95% CI: 1.03-6.42; P = 0.044). Telomere length in tumor tissue and PBMC was not associated with any of the clinical endpoints (Table 2). High TERT levels in tumor tissue were consistently associated with worse prognosis: 5-year OS was 69% in patients with TERT levels < 1318 and 44% in those with TERT levels > 1318 (P = 0.009; Fig. 3E). Corresponding figures for 5-year PFS were 60% and 45% (P = 0.033 Fig. 3D). This difference was likely due to worse regional control in patients with elevated TERT levels (95% vs 63%; P < 0.001 Fig. 3B). At multivariate analysis, high levels of TERT in tumor were significantly associated with a higher risk of regional failure (HR = 5.75, 95% CI: 1.16-28.49; P = 0.032), increased risk of progression (HR:2.12,; 95% CI: 1.00-4.47; P = 0.049) and death (HR:3.53, 95% CI: 1.47-8.52; P = 0.005) ( Table 2). TERT levels in SM and plasma were not associated with any time-to-event variables ( Table 2).

Discussion
Data from this study confirm our previous observation, obtained from an independent series of patients 13 , that short telomeres in histologically normal SM are independently associated with a higher risk of mucosal failure. To investigate whether shortened telomeres in SM are the consequence of greater cell proliferation or are linked to the individual's constitutive telomere length, we estimated telomere length in PBMC. Telomere length in PBMC is considered representative of an individual's constitutive telomere length and correlates strongly with that in cells of different tissues, thus serving as a surrogate marker for other tissues 17 . In this series, while telomere length in PBMC did shorten with age thus showing normal telomeric dynamics, the lengths of telomeres in histologically normal mucosa surrounding UADT cancers did not correlate with patients' ages or with telomere length in www.nature.com/scientificreports www.nature.com/scientificreports/ PBMC. Shortened telomeres in SM could therefore be a consequence of greater cell proliferation and/or environmental and lifestyle-related factors such as tobacco smoking and alcohol consumption. Local tumor recurrence at the primary site can be due to residual cancer cells or to the development of second field tumors. The majority (94%) of patients undergoing primary surgical treatment in this study had clear margins, and none of the four cases with close or involved margins suffered local mucosal failure. This suggests that second field tumors might play a crucial role in mucosal failure. The model of field cancerization has recently been reviewed in molecular terms 18 . According to the model, precancerous fields characterized by genetic alterations due to carcinogen exposure can precede the emergence of synchronous or more frequently metachronous multiple cancer lesions. Part of this at-risk mucosa remains after the primary tumor has been treated and is the site of origin of subsequent lesions 18,19 . A more recent investigation found that the majority of the genomic changes, i.e. copy number www.nature.com/scientificreports www.nature.com/scientificreports/ alterations and point mutations, driving oral carcinogenesis take place in the pre-cancerous condition through gradual casual accumulation rather than as a consequence of a dramatic single event 20 .
Interestingly, patients exposed to tobacco and alcohol had shorter telomeres in histologically normal SM with the association being statistically significant in ever drinkers. Overall, the comparison of telomere length in adjacent mucosa and in PBMC in respect to age indicates that the phenotypically normal mucosa surrounding the tumor constitutes an acquired telomere-shortened epithelial field prone to genetic instability. This hypothesis is supported by findings of previous investigations that observed the presence of TP53 mutations in the field of macroscopically normal mucosa surrounding HNSCC in more than one-third of patients 21 and the generation, due to telomere dysfunction in TP53 mutant mice, of non-reciprocal translocations initiating the neoplastic process and accelerating carcinogenesis 22 .
Furthermore, we found that short telomeres in SM are associated with a worse response to multimodal therapy, including disease progression during treatment. According to some studies 18,19 , the cancer cells of the primary tumor and the genetically unstable cells of the SM belong to a common pre-neoplastic field. Most patients treated with radical surgical resection of the tumor show genetically altered margins upon molecular analysis, and appear to be at high risk of disease progression since the genetically altered mucosa surrounding the primary tumor progressively transforms. Other theories argue that disease progression is due to the acquired migratory capacity of the neoplastic cells through the submucosa or to their excretion in the lumen of a hollow organ (e.g. UADT) with subsequent regrowth in nearby locations 23,24 . Confirmation of these results and determination of their clinical implications might lead to more aggressive therapeutic strategies, more accurate evaluation of the disease response under therapy, and a stricter and more personalized follow-up regime in patients with shorter telomeres in SM. Although there is evidence that telomere function can influence radiosensitivity 15,16 it was not possible to assess this in our clinical cohort because most of patients received (chemo)radiotherapy as upfront treatment, or as adjuvant therapy following surgical treatment and the small number of patients treated by surgery alone did not allow to explore this hypothesis. www.nature.com/scientificreports www.nature.com/scientificreports/ Tumors harboring higher TERT levels were more frequently diagnosed in advanced stage and associated with the involvement of the regional lymph nodes. Moreover, risk of regional failure, progression, and death were independently associated with high levels of TERT thus confirming previous data showing that TERT levels in cancer tissue increase with the aggressiveness of the disease 10,13,25,26 . This association may be attributable to telomerase's non-canonical functions which are implicated in regulating several cancer hallmarks including cell proliferation, resistance to apoptosis, invasion, and metastasis by interaction of TERT with important cancer-related signaling cascades, such as Wnt/β-catenin and NF-kB pathways 10,27 . The activation of NF-kB signaling pathway by TERT stimulates the epithelial-to-mesenchymal transition that provides cancer cells with a more migratory mesenchymal phenotype 12 . High tumor TERT levels also correlate with a worse response to treatment, and it has been demonstrated that TERT confers resistance to different apoptotic stimuli including treatment with chemotherapeutic agents and radiation [28][29][30] . Accordingly, TERT inhibition, through activation of DNA damage response non related to telomere shortening, sensitizes cells to the pro-apoptotic effect of chemotherapeutic drugs 31 . Thus, it is www.nature.com/scientificreports www.nature.com/scientificreports/ conceivable to attribute the negative predictive value of TERT levels to its previously discussed extra-telomeric properties. Furthermore, patients with HPV-driven oropharyngeal cancer had higher levels of TERT in SM. This observation may be related to the topographic restriction of HPV-driven HNSCC to the lymphoepithelial tissue of the palatine tonsils and base of the tongue, potentially characterized by lymphocytes overexpressing TERT 13,32,33 , that with high probability could have been collected when the sampling of the mucosa surrounding the oropharyngeal carcinomas was performed.
Circulating cell-free TERT mRNA levels have been found to be higher in cancer patients with respect to controls, as well as independent markers of tumor response, and prognostic of disease progression in cancers frequently presenting with high tumor mass, e.g. gastric and colorectal cancer 25,34,35 . In the present study, we failed to find an association between plasma TERT levels and outcome variables. This negative result may be due to the low tumor volume and the greater propensity of HNSCC to spread along lymphatic channels rather than hematogenous routes 36 . In a substantial fraction of cases it was not possible to detect specific TERT mRNA in the plasma. It has been proposed that plasma circulating TERT levels could derive from activated lymphocytes and therefore they may be indicative of the induction of tumor immune responses 37 . Further studies are needed to clarify the significance and the role of circulating TERT levels in HNSCC.
In conclusion, high levels of TERT transcripts in cancer cells represent a reliable prognostic marker for identifying HNSCC patients with worse response to treatment and risk of progression. The different behavior of telomere length in adjacent mucosa and PBMC in respect to age suggests that in a subset of patients the phenotypically normal mucosa surrounding the tumor constitutes an acquired telomere-shortened epithelial field prone to genetic instability. These observations may lead to new perspectives in risk stratification, treatment, and surveillance strategies in patients with HNSCC.

Methods
Patients. In total, 101 consecutive newly diagnosed patients with histologically confirmed HNSCC (larynx, oral cavity, oropharynx and hypopharynx) were included in this prospective cohort study from 2012 to 2017. Due to differing epidemiology, etiology, histology and management, patients with nasopharyngeal carcinoma were excluded. Irrespective of HPV status, a multidisciplinary team decided on treatment planning according to TNM staging. Most T1 and T2 HNSCCs were treated with function-preserving surgery or definitive radiotherapy. Most patients with T3 and T4 tumors underwent radical surgery followed by post-operative (chemo)-radiotherapy depending on the presence of adverse features, or upfront chemoradiation. The local institutional review board approved the study protocol ("Comitato Etico Sperimentazione Clinica per le Province di Treviso e Belluno", Italy; ethic vote: 346/AULSS9) and all patients gave their informed consent. We confirm that all experiments were performed in accordance with relevant guidelines and regulations. Along with the blood sample, two solid tissue samples were collected from each patient before treatment, one from a non-necrotic area of the carcinoma, and the other from SM. A UADT fiberoptic evaluation as well as computed tomography (CT) and/or magnetic resonance imaging (MRI) of the primary tumor and the neck were carried out 8-12 weeks after treatment to assess tumor response. The routine follow-up program consisted of locoregional examination including UADT endoscopy at 2-month intervals during the first year, 3-month intervals in the second year, 4-month intervals between the third and fifth year, and every 6 months thereafter.
Tissue samples. Tissues samples from both tumor and SM were obtained before treatment. Surgeons were also requested to biopsy uninvolved mucosa at around 4 cm from the tumor margins depending on anatomical location. Precautions were taken to not contaminate the SM with tumor samples by changing surgical blades each time before cutting tissue. SM tissue was available for 96 patients. Both samples were snap-frozen in liquid nitrogen and stored at −80 °C until analysis. Cryostat sections, 6 microns thick, from each tissue sample were prepared using a 1720 Digital cryostat (Leitz, Germany). One section of each sample was stained with haematoxylin-eosin for histopathology. SM was histologically assessed and, in all cases, no histopathological alterations were found. For each tumor and SM sample, 5 or 6 cryostat sections were collected into two different 2 mL eppendorf tubes and stored at −80 °C and −20 °C respectively, for RNA and DNA extraction. DNA was extracted by the standard phenol/chloroform method, while RNA was extracted with Trizol reagent (Life Technologies, Carlsbad, CA, USA) and reverse-transcribed into cDNA using the SuperScript TM III RNase reverse trancriptase assay (Thermo Fisher Scientific, Waltham, MA, USA), according to the manufacturer's instructions.
Peripheral blood samples. Plasma and PBMC samples were obtained, before treatment, from 6 mL of peripheral blood by Ficoll Paque (GE Healthcare, Chicago, IL, USA) protocol, and stored at −80 °C and −20 °C respectively, until use. DNA for telomere length analysis was extracted from PBMC by the standard phenol/ chloroform method, while the circulating mRNA for the quantification of TERT transcripts was extracted from plasma samples as previously described 34 , with the only modification that in this series of samples we started from 1 ml instead of 500 μL of plasma and all reagent quantities were adjusted accordingly. RNA was reverse transcribed into cDNA using the SuperScript TM III RNase reverse trancriptase assay (Thermo Fisher Scientific) in a final volume of 80 μL, according to the manufacturer's instructions.
Quantification of TERT transcripts in tissue and plasma samples. The level of TERT transcripts was quantified by real-time PCR, as previously described 25 with some modifications. In particular, a new primer pair was designed in order to reduce the length of the amplified product, thus improving detection of the cDNA target sequence 38,39 . Primers AT1 (5′-CGGAAGAGTGTCTGGAGCAA-3′) and AT2b (5′-CGCAGCTGCACCCTCTTCA-3′) were designed on exon 3 and 4 respectively; they bind to nucleotide sequences located upstream of the RT motif 1 on the TERT gene allowing amplification of all TERT transcripts producing an amplified product of 68 bp 34  www.nature.com/scientificreports www.nature.com/scientificreports/ plasma samples, levels of TERT mRNA were also estimated per mL using the conversion factor x 8 as previously described 34 . Telomere length measurement. Relative telomere length was determined on DNA extracted from tissues and PBMC by multiplex quantitative Real-time PCR, as previously described 40,41 , with minor modifications. In particular, each PCR reaction was performed in a final volume of 25 μL, containing 5 μL sample (10 ng DNA) and 20 μL master-mix ready-to-use 1X Light Cycler 480 SYBR Green I (Roche Life Science, Penzberg, BY, Germany), containing 900 nmol/L of each primer. Sequences of telomere and albumin primers are detailed in a previous study 40 . The thermal cycling profile was 15 min at 95 °C, two cycles of 15 s at 94 °C, 15 s at 51 °C, followed by 40 cycles of 15 s at 94 °C, 10 s at 62 °C, 15 s at 74 °C, 10 s at 84 °C, 15 s at 89 °C, with signal acquisition at the end of both the 74 °C and 89 °C steps. After cycling, a melting curve program was run starting with a 95 °C incubation for 1 minute, followed by continuous acquisitions every 0.2 °C for 45 °C to 95 °C (ramping at 0.11 °C/s). A standard curve was generated at each PCR run, consisting of DNA from the RAJI cell line, serially diluted from 10 to 0.41 ng/µl 42 . All DNA samples and reference samples were run in triplicate. LightCycler raw text files were converted using the LC480Conversion free software (http://www.hartfaalcentrum.nl/index.php?main=files&-fileName=LC480Conversion.zip&description=LC480Conversion:%20conversion%20of%20raw%20data%20 from%20LC480&sub=LC480Conversion), and the converted data were analysed using LinRegPCR free software 43 . Telomere length values were calculated as telomere/single-copy gene (T/S) ratio, as previously described 42 . Statistical analysis. Telomere length in tumor, SM, PBMC according to age groups were displayed by beeswarm plots. The association between telomere length and age was tested through the analysis of variance (ANOVA) with contrasts for linear trend. Differences in telomere length and TERT level according to socio-demographic and clinical characteristics were tested with the Kruskall-Wallis test.
The association between telomere lengths, TERT levels, and treatment response was evaluated by calculating the odds ratio (OR) of partial response or disease progression versus complete response through unconditional logistic regression model. OR and 95% corresponding confidence intervals (CI) were adjusted for gender, and potential clinical confounders (namely, cancer site, stage, and surgery).
The impact of telomere length and TERT levels on survival outcomes was evaluated through survival analysis, comparing patients with telomere length and TERT levels under (low) or above (high) tissue-specific median values. For each patient, person-time at risk was computed from the date of last treatment to the event date or the date of last follow-up, whichever came first. Event was defined as cancer reappearance at the UADT site for mucosal control; neck lymph node failure for regional control; distant metastasis for distant control; mucosal, regional or distant recurrence or death for progression-free survival; death for overall survival. The Kaplan-Meier method was used to generate crude survival curves and the log-rank test was used to assess the heterogeneity in time to event in strata of selected covariates 44 , censoring follow-up at 5 years. Hazard ratios (HR) and the corresponding 95% CI were calculated using Cox proportional hazards models 44 , adjusting for gender, age, and potential confounders (namely cancer site, stage, and surgical treatment).

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.