Studies focused on identifying and setting out all the different molecular events involved in the early stages of the immortalization process that refine the prognostic value of known markers are needed. The TERT locus is a critical vulnerability site for tumor progression, dedifferentiation, and aggressiveness in thyroid carcinogenesis. In addition to TPM, the alterations that may result in TERT re-expression and telomerase activation in thyroid cancer cells include genomic amplification as well as structural alterations at the TERT locus, chromatin remodeling, or hypermethylation events upstream of the TERT transcriptional start site [15]. Most of those mechanisms, which might significantly impact on telomerase function in thyroid cancer cells, are not fully understood because most of the research efforts so far have concentrated on the analysis of TPM [16,17,18,19,20]. Since 2014, when the TCGA research network approached the genomic characterization of PTCs [5], very few original research studies have focused on investigating the prevalence of somatic TERT CNV (copy number gains or amplification) as a mechanism of telomerase reactivation in PTCs [7,8,9].

In this study, we focused on TERT amplification as an alternative mechanism to TPM for TERT re-expression and telomerase activation. The overall prevalence of TERT amplification found among PTCs, PDCs, and ATCs was 17%, 20%, and 17%, respectively. In agreement with the theory that considers the molecular alterations involved in TERT re-expression and telomerase activation as events associated with advanced stage, clinically aggressive carcinomas, in our study, TERT amplification was much more frequent in PTCs with DMs. An 86% of the PTCs bearing TERT amplification were from the subset of patients with DMs. The prevalence among PTCs with paired LNMs that did not develop DMs at diagnosis and/or during the follow-up was similar to that seen in the conventional, low/intermediate risk PTCs without DMs genotyped by the TCGA research network (4.7% this study vs. 4.4% TCGA study) [5].

The study of 13 thyroid cancer cell lines revealed, for the first time, in vitro study models that reproduced the results found in tumor samples.

In the TCGA study, arm level alterations occurred more frequently in FV-PTCs than in CL-PTCs (P < 0.008). Unsupervised clustering of chromosomal arm-level alterations defined 4 distinct groups of PTCs, one of which was characterized by a higher frequency of focal somatic CNV (gains and losses) and was enriched for FV-PTCs. Copy gains/amplifications at 5p (TERT) were found in 3.4% of the CL-PTCs, 9.5% of the FV-PTCs, and 2.85% of the tall cell PTCs analyzed by the TCGA [5]. Likewise, we have also found that TERT amplification was more frequent among the FV-PTCs (21.05% of the total cases; 57.14% of the cases with amplification. All 4 cases were infiltrative FV-PTCs) than among CL-PTCs (13.63% of the total cases; 43% of the cases with amplification). In our series, however, we did not see a significant correlation with the follicular variant tumor subtype. In 2016, Yoo SK and colleagues also found a higher percentage of somatic arm-level CNV in FV-PTCs than in CL-PTCs [21]. In the past, it has been hypothesized that CNV and chromosomal instability may drive FV-PTCs and pathogenesis, with CNV leading to a different tumor histology [22].

To verify the concurrent or mutually exclusive nature of TERT amplification and TPM and the alleged association of TERT amplification with other oncogenic drivers in thyroid carcinogenesis, we investigated in the same cohort of tumors the presence of TPM, BRAF, H-RAS, K-RAS, N-RAS, and PIK3CA mutations. In contrast with Gupta’s findings [8] and similarly to results of other authors [6, 7, 9, 10], in our study, TERT amplification and TPM co-occurred in 15%, 0%, and 9% of the PTCs, PDCs, and ATCs, respectively. An 83% of the PTCs bearing both events were from the subset of patients with DMs. In PTCs, but not in PDCs and ATCs, TERT amplification and TPM were significantly correlated (P = 0.0313). It has been shown that TERT expression is significantly higher in tumors with TERT amplification than in tumors without TERT amplification (P = 0.04), as well as that the combination of TERT amplification and TPM further increases the expression [5]. It appears that the two events might cooperate in TERT reactivation and thyroid cancer progression. Barthel et al. showed that the highest telomerase activity is found in tumors with TERT amplification [2]. When genome data sets of different types of cancers (lung squamous cell carcinoma, bladder urothelial carcinoma, metastatic melanoma, and hepatocellular carcinoma), with information on TERT mRNA expression and copy number alterations, are uploaded to cBioPortal (http://cbioportal.org), we see that TERT amplification is associated with much higher expression of TERT than TERT copy number gains (see “online resource 3” for detailed information on TERT mRNA expression analyses in different tumors using the web resource cBioPortal for exploring, visualizing, and analyzing multidimensional cancer genomics data sets). In 2021, Gupta S et al. also corroborated that those thyroid tumors with genomic TERT amplification and rearrangements exhibited statistically significant higher increases in TERT expression than tumors with TPM [8]. In our series, the lack of frozen tissue from any of the samples analyzed hampered the correlation between TERT amplification and TERT mRNA expression levels. Considering the different analyses we carried out in cBioPortal and the rest of the research studies mentioned, one would expect that the cases with TERT amplification in our case series would behave with respect to TERT expression in a similar way to those cases reported in the literature that exhibited only TERT amplification or TERT amplification plus TPM. On the other hand, TERT immunohistochemistry (IHQ) or measurement of protein expression did not work properly (data not shown). Curiously, in a study by Paulson et al. in thyroid cancer, no correlation between TERT mRNA expression and TERT IHQ could be demonstrated [23]. Although several studies have attempted to evaluate TERT protein expression using IHQ, this has been controversial. The efforts have been hindered by poor reproducibility, unexpected patterns of subcellular localization, and documented cross-reactivity with other proteins [8, 24, 25]. Thus far, TERT IHQ does not seem to be a useful clinical tool for prognostication [23, 26]. In PTCs, no correlation has been reported between TERT IHQ and clinical-pathological traits [26].

Coexistence of TERT amplification and BRAF mutations was seen in 15%, 7%, and 9% of the PTCs, PDCs, and ATCs, respectively. Only in PTCs was found a trend of correlation between TERT amplification and BRAF mutations (P = 0.0994). Coexistence of TERT amplification and RAS mutations was seen in 7%, 7%, and 9% of the PTCs, PDCs, and ATCs, respectively. When the three histotypes investigated (PTC, PDC, and ATC) were analyzed together, a trend for an association between TERT amplification and RAS mutations appeared (P = 0.1166). Coexistence of TERT amplification and PIK3CA mutations was seen in 2%, 0%, and 11% of the PTCs, PDCs, and ATCs, respectively. TERT amplification and PIK3CA mutations were significantly correlated (P = 0.0272) in ATC.

While in the TCGA study somatic CNV were significantly enriched in cases with no driver mutation, suggesting that somatic CNV may also drive PTC, in our casuistry only three cases (2 PDCs and 1 ATC) did not exhibit any of the other genetic drivers screened. Most of the cases had 2 or 3 associated mutational events. A putative explanation for this difference is that while in the TCGA study were mainly characterized conventional, low/intermediate-risk PTCs without DMs (98,4%), in our casuistry, have been mainly characterized advanced stage, clinically aggressive cancers (77%). It is known that advanced stage, clinically aggressive tumors are prone to genomic instability and accumulation of different genetic events. Nonetheless, the 5 PTCs with aggressive histology genotyped in the TCGA study revealed, as did the PTCs with DMs analyzed in this study, a much higher mutational burden than conventional low/intermediate-risk PTCs [5]. When the pTs grow slowly, the growth advantage of additional driver mutations is larger, subclones greatly expand, and the rate of mutations in driver genes within the pTs before their clinical detection increases.

Molecular heterogeneity was found in 60% of the PTCs, 60% of the PDCs, and 66% of the ATCs genotyped (see Fig. 1 and Fig. 2). More than half of the PTCs (67%) exhibiting molecular heterogeneity were PTCs with DMs. Two oncogenes were concurrently activated in 67%, 33%, and 58% of the PTCs, PDCs, and ATCs, respectively. Three oncogenes coexisted in 28%, 67%, and 32% of the PTCs, PDCs, and ATCs, respectively. Four oncogenes activated were seen in 6%, 0%, and 11% of the PTCs, PDCs, and ATCs, respectively. TERT activation may predispose, through the induction of genomic and chromosomal instability, to the acquisition of secondary genetic events. Additional molecular alterations, which, in turn, may activate signaling pathways that account for the aggressiveness of some PTCs.

Driver gene mutation heterogeneity within primary PTCs and matched DMs is commonly ignored in the clinical setting leading to an inadequate risk-based stratification of patients, improper clinical surveillance, and erroneous therapeutical planning.

To date, none of the few published studies on TERT amplification in PTCs has addressed the clonal or non-clonal nature of TERT activation by TERT amplification or by TPM, in the case of having analyzed both events.

In our study, we observed that in PTCs TERT amplification was a subclonal event. Although in more than half (57%) of the PTCs showing variations in TERT gene dosage was found an increase in TERT copy number in all the tumor samples analyzed in each case, not in all of the areas genotyped in those cases the increase reached the established amplification threshold. Watkins TB et al. have demonstrated that continuous chromosomal instability results in pervasive somatic CNV heterogeneity. Using a multi-sample phasing in the analysis of somatic CNV across 22 tumor types showed that focal TERT amplifications were frequently subclonal [27]. In PDCs and ATCs, when concurrent better differentiated areas within the pT were genotyped, we saw that TERT amplification appeared to show a preference for the phenotypically less differentiated areas.

In our study, TPM were found to be clonal in half of the mutated PTCs. When clonality was assessed within the pTs of the mutated PTCs, TPM were clonal in more than half of the cases (63%), mostly advanced PTCs with DMs. Within the LNMs and DMs, clonality reached levels of 50% and 75%, respectively. Regardless of whether the DMs were synchronous or metachronous, TPM tended to be present in all of the different DMs screened. The increase in the rate of TPM clonality at DMs appears to indicate that the TERT-mutated bearing cell clusters that extravasate and reach the metastatic sites do not experience serious constraints for survival and homing. It might occur a clonal overgrowth of cells with TPM at DMs, meaning that cells with TPM, which are known to have a growth advantage, overgrow all other cells. Among PDCs and ATCs, clonality was only investigated within pTs, observing in both histotypes that TPM were clonal. Our data suggest that there is a trend towards clonality of TPM with tumor progression, dedifferentiation, and clinical aggressiveness (DMs), which is consistent with previous findings reported by Landa et al. on TPM prevalence and clonality in thyroid cancer [28]. Landa et al. found that TPM were scarce and subclonal in conventional PTCs and highly prevalent and clonal in the more aggressive types of thyroid cancer (PDC and ATC). Similarly, the genotyping of 355 PTCs by Liang et al. revealed that TPM occurred in a subclonal manner [29].

Our results regarding the clonal nature of TERT activation events clearly illustrate that when genotyping is restricted to only one area, primarily of pT or LNM and less crucially of DM, it can occur that subclonal TERT amplification or TPM may evade detection. It has been shown, in different tumor types, that genetic analyses from single biopsies may lead to an underestimation of the complex mutational portraits that characterize different advance stage tumors [30,31,32,33]. Another detail to have in consideration when analyzing the putative subclonality of a mutational event is the sensitivity of the methodology applied to detect mutant alleles. It has been demonstrated, in PTCs and follicular thyroid tumors of uncertain malignant potential, that digital droplet PCR (ddPCR) has a higher sensitivity for detecting TPM in samples that have very low mutant allele frequencies [34, 35]. The latter means that we might have missed some TPM and, thus, the clonality observed in the whole series of PTC could be higher than the reported of 50%.

To know the clonal status of different genetic events within a particular tumor it is crucial for the success of personalized medicine. Targeting of subclonal events will be certainly insufficient to prevent tumor progression. Indeed, inhibition of subclonal alterations will most probably only cause the outgrowth of other untargeted mutated cancer cells that drive tumor progression, as well as the failure of targeted therapies and a fatal outcome [36]. The commonest is that clinicians make a decision in favor a particular targeted therapy on the basis of a single biopsy of the pT, which may not reflect the clonal status of the targeted gene, a mistaken strategy that facilitates the appearance of therapy resistance.

Heretofore, our study has screened the largest series of primary PTCs with paired LNMs (30 cases) and primary PTCs with matching synchronous and/or metachronous DMs (20 cases). Near 70% of the PTCs showing TERT amplification at the pT revealed also TERT amplification in at least one of the metastatic niches genotyped. TPM were found to spread with metastatic PTC cells to LNMs or DMs in 73% and 67% of the cases, respectively. Apparently de novo mutations at LNMs or DMs were seen in 18% and 33% of the PTCs, respectively. In most of the mutated cases, the metastatic PTC cells maintained their primary mutational traits after metastases development, with limited influence of the metastatic niche. In some of the mutated cases, however, regional pressures, presumably exerted by the homing microenvironment at metastases, could have led to the appearance of metastatic tumor cells bearing advantageous de novo/private mutations. The more common C228T TPM exhibited a greater propensity to spread from the pT to metastases, while the less common C250T TPM seemingly originated more frequently de novo, at the metastases. It is tempting to speculate that the C250T TPM is prone to emerge in metastatic milieus, contributing not only to shape the metastases but also the appearance of resistance to targeted therapies. It could also be that the mutations considered as de novo mutations in LNMs and DMs are not such. We cannot rule out the possibility of having lost subclonal TPM in pTs. Mutations present in a small number of pT cells may evade detection. It has been shown in PTCs that TPM in samples that harbor very low mutant allele frequencies can be missed by sanger sequencing [35].

TERT activation by TERT amplification and/or TPM spread from the better differentiated area to the less differentiated area within the pT in two-thirds of the PDCs and 100% of the ATCs. Despite the limited number of PDCs and ATCs in which it was possible to investigate the transfer of the TERT activation mechanism from a better differentiated area to a less differentiated area within the pT, it is tempting to hypothesize that once TERT is activated it is prone to evolve with tumor cell dedifferentiation. The dissemination of TERT amplification with cellular dedifferentiation represents additional molecular evidence for a stepwise progression from PTC to PDC and ATC within a multistage genetic model of thyroid follicular cell tumorigenesis, but does not necessarily mean that it is the driving force that underlies histological dedifferentiation. The overall prevalence of TERT amplification in PDCs and ATCs, seen in this study and previous studies, is lower or similar to that demonstrated for other oncogenes, which implies that it is not a dominant event in the pathogenesis of PDCs and ATCs.

Up till now, none of the few studies performed on TERT amplification as a mechanism of TERT re-expression and telomerase activation in thyroid cancer [5,6,7,8,9] has appraised the impact of TERT amplification on the clinical course (recurrence and survival) of PTC patients. Only Paulsson et al. have assessed the impact of TERT aberrancies, including TERT copy number gains, in tumor-related relapse, but the tumors investigated were FTCs not PTCs [6]. To the best of our knowledge, this is the first study that evaluates the relationship between TERT amplification in PTCs and clinical-pathological parameters of poor prognosis, recurrence, and survival. Some of the previous studies concur in considering TERT amplification, like TPM, as a late event in thyroid carcinogenesis, more common in advanced tumor stages. None of those studies has proved, however, a statistically significant correlation between TERT amplification and the patient’s tumor stage. Importantly, some of these studies also agree on the need to analyze larger series of carcinomas in advanced stages, aimed at defining the impact that pathogenic alterations of TERT, different from TPM, have on the prognosis of patients [7,8,9].

Our analysis of PTC patients reveals that TERT amplification, as shown in Table 2, has an enormous impact on the clinical course of PTC patients. The statistically significant associations that we found with vascular invasion, DMs at diagnosis and/or during follow-up, and metachronous DMs, which develop during the patient’s follow-up, are those that can have a greatest impact on patient’s prognosis. Based on our results, TERT amplification is a major determinant of the metastatic capability of PTCs, increasing the risk of DMs at diagnosis or during the follow-up. De facto, when TERT amplification was present in PTC cells, a significant relationship was observed with tumor stage at diagnosis, stage III/IV at last follow-up, and a DOD patient status. Our findings are consistent with reported data showing that TERT, independently of its function in maintaining telomere length, participates in the activation of the epithelial-mesenchymal transition (EMT), which implies the induction in cells of migratory and invasive capacities. TERT interacts with β-catenin (one of the EMT-associated transcription factors) and together associate with mesenchymal marker promoters to drive their expression [37]. Moreover, it has been demonstrated in highly metastatic PTCs that CDH6 expression, a class II cadherin aberrantly reactivated in cancer, is restricted to EMT cells that also exhibit a higher incidence of TERT amplification, a finding that raises the hypothesis of a putative functional connection of both events in EMT activation and metastatic spreading [38, 39]. Similar to what we have previously reported concerning the role of BRAF and RAS mutations in the spread and homing of mutated BRAF or RAS cells in lymph nodes [40, 41], in this study, we also saw that TERT amplification disseminated with PTC cells, but did not drive the development of LNMs in PTCs. Metastatic tumor expansion into lymph nodes can occur independently of TERT amplification, a finding consistent with the lack of correlation of TERT amplification with tumor multifocality, which is known to increase the likelihood of developing LNMs in PTCs. Importantly, the concurrence of other oncogenic drivers (TPM, BRAF, or RAS) and TERT amplification does not improve the correlations of prognostic relevance observed with TERT amplification per se. On the contrary, some of the observed correlations become just a tendency towards an association or worsened notably. Only the conjunction of TERT amplification and TPM determined that the tendency towards the association observed between TERT amplification and age became statistically significant.

Our data clearly point to a greater impact of TERT amplification than TPM on the prognosis of PTCs. The only prognostic traits with which TPM were significantly correlated were the patient’s age, tumor stage, and DOD status. In contrast to what was observed with TERT amplification, the concurrence of TPM with TERT amplification or RAS mutations results in the appearance of statistically significant associations with clinical-pathological traits of poor prognosis (vascular invasion, DMs at diagnosis and/or during follow-up, and metachronous DMs) that with TPM per se were not significant. The latter does not apply to the coexistence of TPM and BRAF mutations, which even leads to the loss of the correlation found between TPM and DOD status.

According to what we and others have previously reported, RAS mutations also have a greater impact on the prognosis of patients with PTC than BRAF mutations [40]. While BRAF mutations were found to correlate only with the presence of focal tall cell futures and the development of tumor recurrences, RAS mutations, as formerly shown [40], demonstrated to be significantly associated to age, DMs at diagnosis and/or during the follow-up, metachronous DMs, developed during the follow-up, and DOD patient status.

See “online resource 4” for discussion of impact of BRAF mutations in recurrence-free survival and risk of death of disease in PTC patients.

We demonstrate for the first time that TERT amplification is associated with a lower probability of recurrence-free survival and a greater risk of death of disease in PTC patients, findings that are consistent with the statistically significant correlation found between TERT amplification and the development of metachronous DMs, during patient follow-up. A relationship with tumor recurrence and shorter survival that has been previously reported in different tumor types [3, 42,43,44]. In our study, we show that TERT amplification is a better predictor of tumor relapse than TPM, extrathyroidal extension, tumor multifocality or tumor size ≥ 5 cm, DMs at diagnosis, and stage at diagnosis I–II vs. III–IV. Furthermore, TERT amplification predicts tumor relapse independently of other variables that also exhibit predictive value such as the presence of foci of infiltrative insulae of tumor cells, surrounded by a desmoplastic reaction at the advancing edge of the tumor, presence of areas of focal tall cell appearance, age ≥ 55 years old, male sex, vascular invasion, LNMs at diagnosis, and BRAF mutations. Likewise, our findings also evidence that TERT amplification is a better predictor of tumor-related death than TPM, male sex, tumor multifocality or tumor size ≥ 5 cm, vascular invasion, DMs at diagnosis and/or follow-up, and stage at diagnosis I–II vs. III–IV. Additionally, we show that TERT amplification is able to prognosticate lower survival independently of other variables that also display predictive value such as age ≥ 55 years old, extrathyroidal extension, and BRAF mutations. In contrast to TERT amplification, the presence of BRAF mutations increases the likelihood of recurrence-free probability and decreases the risk of tumor-related death. It has a protective effect.

Beyond the two most common TPM, which selectively recruit the ETS transcription factor GABP to activate TERT, the mechanism of telomerase reactivation in case of TERT amplification in thyroid cancer is poorly characterized. Mechanistic studies have been approached in bladder and glioblastomas cell lines bearing constructs that mimic TERT duplications containing a de novo native ETS motif for the GABP transcription factor complex. ETS-harboring TERT duplications were first reported in PTC and PDC [7] and shortly after in glioblastomas [