Intratumoral Genetic Heterogeneity in Papillary Thyroid Cancer: Occurrence and Clinical Significance

Intratumoral heterogeneity (ITH) refers to a subclonal genetic diversity observed within a tumor. ITH is the consequence of genetic instability and accumulation of genetic alterations, two mechanisms involved in the progression from an early tumor stage to a more aggressive cancer. While this process is widely accepted, the ITH of early stage papillary thyroid carcinoma (PTC) is debated. By different genetic analysis, several authors reported the frequent occurrence of PTCs composed of both tumor cells with and without RET/PTC or BRAFV600E genetic alterations. While these data, and the report of discrepancies in the genetic pattern between metastases and the primary tumor, demonstrate the existence of ITH in PTC, its extension and biological significance is debated. The ITH takes on a great significance when involves oncogenes, such as RET rearrangements and BRAFV600E as it calls into question their role of driver genes. ITH is also predicted to play a major clinical role as it could have a significant impact on prognosis and on the response to targeted therapy. In this review, we analyzed several data indicating that ITH is not a marginal event, occurring in PTC at any step of development, and suggesting the existence of unknown genetic or epigenetic alterations that still need to be identified.


Introduction
Genomic analysis of cancer samples reveals a complex mutational landscape with vast intertumor and intratumoral heterogeneity. Intertumoral heterogeneity refers to genetic and phenotypic variants occurring among individuals with the same tumor type. Intratumoral heterogeneity (ITH) refers to a subclonal diversity that may be observed within a tumor lesion. Intratumoral genetic heterogeneity is a paradigm of carcinogenesis, a process that transforms a tumor into a more aggressive cancer through gain of genetic alterations. This process transforms a clonal neoplasm into a mass of genetically different subclones that may be intermixed or spatially separated within the neoplastic tissue. Tumor subclones are characterized by differential gene expression due to both genetic and epigenetic heterogeneity, such as chromosome copy number variations, point mutations, genes rearrangements or epigenetic modifications that result in phenotypic diversity and intercellular heterogeneity. Interestingly, this intercellular heterogeneity, which may promote the development of new subclones, is empowered by genomic instability, which is in turn influenced by cancer treatment [1].
In human tumors, ITH has been documented by Software Inference methods able to infer evolutionary relationships between clonal subpopulations based on variant allele frequencies of point mutations and taking into account copy number alterations at the mutated loci [2,3]. According to these analysis, ITH is highly variable among tumors of different type. For instance, melanomas are involved, which are frequently mutually exclusive, compared to other cancers [22], such as lung cancer and melanoma, where multiple oncogenes are frequently found to be altered in the same tumor at early stages [5]. As an example, subclonal BRAF V600E , the most frequent mutation in melanoma, was described in several reports. In particular, single cells genotyping showed BRAF V600E heterogeneity within metastatic primary melanomas [23], in melanoma metastases [24], and among circulating tumor cells [25]. Similarly, subclonal RAS mutations were described in melanomas, and N-RAS and BRAF activating mutations were demonstrated to coexist in the same melanoma in different cells [26].
Unlike melanoma, PTC is considered to be largely homogeneous, so that a subtype classification was proposed according to the mutation detected, i.e., BRAF-like and RAS-like PTC [22]. Extended analysis of large series of tumors demonstrated a relatively low overall density of somatic mutations that is believed to be the biological basis for the indolent clinical behavior of PTC. Technical issues can partly explain the scanty data on ITH in PTC: the genetic pattern of PTCs was mostly investigated by the low sensitive sequencing Sanger method and was not to identify mutations present with a low allelic frequency [27]. Thyroid tumors consist of neoplastic cells intermingled irregularly with normal (connective tissue and vessels) and reactive (stromal and immune) cells, and the ratio between these components may vary largely between tumors [28]. Thus, the neoplastic cells content and the allelic frequency of the mutated oncogene in a given sample can be extremely low and below the sensitive threshold of many analytical methods. Accordingly, in PTC the allelic frequency must be normalized for the percentage of tumor cells in the sample, a measurement not always easy to perform [22,29]. Indeed, the number of the studies included in the present review which report data on the genetic characterization and/or ITH in TC, not evaluating tumor purity [27, are significantly more numerous than those which normalized the data for the percentage of tumor cells [7,20,22,29,[53][54][55]. Thus, data on ITH obtained without considering the purity of the tumor must be considered with caution. Moreover, it should be underlined that a rigorous method to establish the genetic heterogeneity and clonality of cancer should include the evaluation of copy number alteration, too. Recent data from the pan genomic characterization of a synchronous FTC, PDTC and ATC showed that the cancer cells fraction determination (CCF), which denotes the proportion of cells among all cancer cells carrying a specific genetic aberrancy, allows precisely establishing the clonal composition of the tumors during the dedifferentiation process [19].
More recent methods of genetic analysis, i.e., pyrosequencing, allele-specific locked nucleic acid PCR, and next-generation sequencing, made possible a deeper and quantitative analysis of the mutational status of tumor samples, providing data in favor of ITH in PTC. Nevertheless, the existence and the relevance of subclonality in PTC is still debated since discordant evidence is available to date [56]. Moreover, the contribution of intratumoral heterogeneity to thyroid metastatic cancers and the clonal relationships between the primary thyroid tumor and lymph node or distant metastases are still unknown.

Evidence in Favor of ITH in PTC
Much evidence has been reported supporting the occurrence of ITH in PTC, either in early or advances stages of progression (Table 1).

Heterogeneous Presence of a Mutation Documented by Genetic Analysis
The first evidence of ITH in PTC came from studies evaluating the presence of RET/PTC rearrangements. The distribution of RET fusions was investigated by means of different approaches, demonstrating to vary in sporadic PTC or in post-Chernobyl PTC cases. The analysis of RET rearrangements by interphase fluorescence in situ hybridization (FISH) in 29 adult and 13 childhood post-Chernobyl PTCs unveiled that in all positive cases (23 and 10, respectively), the tumors were composed of a mixture of cells with and without RET rearrangements [30,31]. This ITH was further confirmed by a different research group that analyzed by FISH 14 RET/PTC positive PTC, finding nine cases with 50%-86% positive cells and five cases with 17%-35% positive cells [32]. High level of recombinant RET/PTC mRNA, a finding that the authors considered compatible with a clonal occurrence of the recombination, was observed only in 46% of RET rearrangements-positive adult PTC [33]. Interestingly, immunohistochemistry and reverse transcriptase-polymerase chain reaction (RT-PCR) analyses performed on RNA extracted after laser capture microdissection, and FISH experiments demonstrated the subclonal occurrence of RET/PTC rearrangements not only in PTC but also in hyperplastic or adenomatous nodule and even in scattered thyroid cells in Hashimoto's thyroiditis [33,34]. In the study by Zhu et al. different detection methods with different sensitivity (standard-and high-sensitivity RT-PCR, real-time Light Cycler RT-PCR, Southern blot analysis, and FISH) demonstrated the subclonal or non-clonal occurrence of RET/PTC-1 and -2 in 17 of 65 (26%) PTC, while the clonal occurrence was demonstrated only in 9 (14%) tumors [32].
In following years, ITH of BRAF V600E was demonstrated in PTC by different molecular techniques ( Table 2). By pyrosequencing analysis, the subclonal or even oligoclonal occurrence of BRAF V600E mutation was found to be more frequent than the clonal occurrence [36,57]. BRAF V600E was demonstrated in more than 45% of alleles only in a minority of cases, indicating that BRAF V600E mutation is frequently an oligoclonal event [57]. This data was confirmed by Gandolfi et al. [7] who showed by pyrosequencing in 58 BRAF V600E -positive PTC, an average mutated allele percentage Cancers 2020, 12, 383 5 of 15 of 27.44%, with a range between 7.50% and 49.80%. Further demonstration that both BRAF V600E positive and negative cells can coexist in classic PTC was achieved by pyrosequencing of 264 manually microdissected BRAF V600E -positive tumors in which the mutant allelic frequency ranged 8%-41% of the total BRAF alleles (median, 20%) [37]. Based on these studies, for both RET/PTC and BRAF V600E the subclonal occurrence in PTC appears to be something more than a rare event. The demonstration that the PTC tumor mass frequently consists of a mixture of few cells bearing mutant BRAF and more abundant tumor cells bearing wild-type BRAF was further confirmed after normalization for the percentage of tumor cells, and by the analysis of single cells obtained by laser capture [57]. De Biase et al. applied the allele-specific locked nucleic acid PCR to 155 PTC to determine the presence of BRAF V600E and the allelic frequency after subtraction of non-tumor cells [29]. They observed that 10.6% PTCs displayed < 30% of BRAF V600E neoplastic cells, and 45.9% PTCs displayed 30-80% of BRAF V600E neoplastic cells. Overall, the 63.8% of PTCs had less than 80% of mutated cells. The heterogeneous distribution of BRAF V600E in PTC, indicating subclonality or even oligoclonality, was confirmed in subsequent studies by means of next generation sequencing (NGS). The analysis of 30 BRAF V600E -positive PTC by 454 NGS, revealed a mean and median of BRAF V600E -positive neoplastic cells of 72.3% and 83%, respectively [29]. In another study, after the exclusion of non-tumoral cells by means of a morphometric analysis, 24% of 49 BRAF V600E positive PTC were found to be subclonal by Ion Torrent-NGS [53]. These agreeing findings obtained in different laboratories by means of different methods, led to hypothesize that this mutation is not always the first transforming genetic event and that it can be a secondary event in PTC tumorigenesis [38,39]. More recently, the analysis of the ITH of PTC was extended to TERT and RAS genes. After normalization for tumoral cell content, MassARRAY genotyping confirmed the finding of the mutations in these two genes at low allelic frequencies in some samples, consistent with their presence in a small subset of cancer cells [20,24]. Recent data obtained by a multi-region WES approach on 257 PTC tumor tissues showed the presence of a subclonal driver alteration in 29% of tumors [58].  [66]. The heterogeneous staining described in some melanoma samples in the first report, was imputed to necrosis, though in subsequent studies this technical issue was no more reported and this antibody was considered a valid tool to investigate the intratumoral distribution of BRAF V600E [66]. In a series of 85 PTCs analyzed by immunohistochemistry (IHC) with the VE1 antibody, 37 cases (43.5%) displayed more than 80% of stained cells, 39 cases had 30%-80% (45.9%), and nine had less than 30% (10.6%) stained cells [29]. In a different laboratory, when stained with the same antibody, immunoreactive and non-immunoreactive PTC cells were clustered separately or were intermingled in the primary lesions and in the corresponding metastatic lymph nodes [7]. These data indicated that as for the primary lesions, the matched lymph nodes where heterogeneous for the BRAF V600E mutation. Contrasting data come from Ghossein et al. who found a homogeneous staining in 13/14 PTCs, in 3/3 poorly differentiated TCs, and in 12/14 anaplastic TCs, supporting the concept that the BRAF V600E mutation is a clonal event in the majority of TCs while ITH is a rare occurrence [67]. IHC with VE1 was employed in many studies to determine its reliability as diagnostic tool, though the issue of heterogeneity has not been addressed and, currently, the tumor is considered positive when a significant percentage of tumoral cells is stained.

Presence of Concomitant Mutations
In recent years, also thanks to advances in NGS technology, it has become evident that multiple mutations can be concomitantly present in the same tumor [68,69]. In the context of PTC, co-occurrence of mutations, such as BRAF V600E , TERT, RET/PTC and H4/PTEN has been frequently documented, indicating that these genetic alterations might coexist in the same tumor [40,[42][43][44][45][46][47]. Dual mutation of BRAF V600E and RET/PTC and of BRAF V600E and TERT promoter [37,48,53,54,57] and of point mutations and fusions [20] were found to occur in up to 20% PTCs. Although sporadically, concomitant occurrence of different RAS mutations or mutations of different RAS isoforms or concomitant RAS mutations and RET/PTC rearrangements were reported [49,50].

Discordant Mutational Status between Primary Site and Metastases
The multistep/multigene model with a progressive acquisition of new genetic defects is a paradigm of carcinogenesis. A discordant mutational status between the primary site and the metastases supports this mechanism and it was recorded in several human tumors (including lung, melanoma, colorectal, gastric, breast) [4]. In TC, discordant patterns were reported for BRAF V600E , TERT, RAS and other mutations [27,39,40,51,52,[58][59][60][61][62][63][64][65]. Data related to PTC are reviewed in Table 3. In particular, the BRAF mutational status between primary tumor and metastases, mainly loco regional, was found to be discordant in up to 50% of cases, by different techniques [39,40,51,52,59,61,62]. In a recent study, it was shown that 14/27 TERT mutated primary tumors had wild-type TERT lymph node or distant metastases [52]. The loss in the metastases of the BRAF or TERT mutation present in the primary site is a very unlikely occurrence, hence these data strongly support the ITH of the primary tumors. Additional findings concerning ITH, and clonal relationship between primary tumor and metastases come from a case report of an aggressive PTC with matched lymphnode and a pleural metastasis. By the analysis of single nucleotide variants, gene fusions, and loss of heterozygosity, the authors showed that some of the genetic alterations were ubiquitously detected in all the tumor samples from the patient and others were detected only in some tumor areas, indicating the presence of several subclones in the neoplastic tissues. Interestingly, the two selected areas of the primary tumor and the two selected areas of on regional metastasis presented similar genetic profiles, whereas the two selected areas of another regional metastasis had more divergent mutations and fusions [55]. Striking heterogeneity was also observed between paired primary tumors and metastases in studies done by means of NGS and WES studies [58,60,63,64]. The contribution of ITH to thyroid metastatic cancers still needs to be more studied and clarified, though the finding of genetic heterogenous PTCs allows proposing hypotheses related to thyroid oncogenesis and progression ( Figure 1).
(A) A known mutation occurs in the thyroid follicular cell and leads to the development of the tumor. Thereafter, the mutation is clonally distributed in all the tumor cells at the primary site and propagated to all the metastases developed during tumor progression. Primary tumor and metastases are clonal and the driver gene is detectable in both. This scenario is likely to be extremely frequent for PTC.
(B) The tumor is established by the transformation of a thyroid cell by a genetic driver (known or unknown). A second genetic event is acquired and transmitted to a subset of tumor cells at the primary site (sub-clonal distribution). The second genetic event can occur either in the same cell or in a different cell. This heterogeneous pattern can lead to the development of metastases with different genetic assets, namely cells with a double oncogenic expression or cells with monogenic expression intermingled with cells with a double oncogenic expression. A) A known mutation occurs in the thyroid follicular cell and leads to the development of the tumor. Thereafter, the mutation is clonally distributed in all the tumor cells at the primary site and propagated to all the metastases developed during tumor progression. Primary tumor and metastases are clonal and the driver gene is detectable in both. This scenario is likely to be extremely frequent for PTC.
B) The tumor is established by the transformation of a thyroid cell by a genetic driver (known or unknown). A second genetic event is acquired and transmitted to a subset of tumor cells at the primary site (sub-clonal distribution). The second genetic event can occur either in the same cell or in a different cell. This heterogeneous pattern can lead to the development of metastases with different genetic assets, namely cells with a double oncogenic expression or cells with monogenic expression intermingled with cells with a double oncogenic expression. C) A second genetic event is acquired at the metastatic site, either in the same cell or in different cells. The acquisition of a second event is predicted to increase the growth potential leading to the development of a clinically evident metastasis.

Evidence Limiting the Impact of ITH In PTC
While large data support the IHT in PTC, its biological significance and clinical impact is debated. In some studies, the occurrence of ITH has been called into question or it is considered a limited or extremely limited phenomenon in PTC. With the Cancer Genome Atlas (TCGA) project, whole genome DNA of a large PTC cohort was examined by NGS [22]. This study reported somatic mutations in 83% and gene fusions in 13% of cases, mostly affecting the RAS/RAF/MAPK pathway. With very few exceptions, all mutations had an allelic frequency below 50%. However, subtracting the non-cancer cells estimated by a computational method that uses a pre-computed statistical models of recurrence cancer karyotypes [70], the major driver mutations documented (BRAF V600E , RAS mutations, and RET/PTC) were present in the majority of tumor cells with only very few exceptions. These findings led the authors to conclude that the tumors were largely clonal and that oligoclonality or polyclonality with respect to these oncogenes is a phenomenon limited to a few PTC

Evidence Limiting the Impact of ITH In PTC
While large data support the IHT in PTC, its biological significance and clinical impact is debated. In some studies, the occurrence of ITH has been called into question or it is considered a limited or extremely limited phenomenon in PTC. With the Cancer Genome Atlas (TCGA) project, whole genome DNA of a large PTC cohort was examined by NGS [22]. This study reported somatic mutations in 83% and gene fusions in 13% of cases, mostly affecting the RAS/RAF/MAPK pathway. With very few exceptions, all mutations had an allelic frequency below 50%. However, subtracting the non-cancer cells estimated by a computational method that uses a pre-computed statistical models of recurrence cancer karyotypes [70], the major driver mutations documented (BRAF V600E , RAS mutations, and RET/PTC) were present in the majority of tumor cells with only very few exceptions. These findings led the authors to conclude that the tumors were largely clonal and that oligoclonality or polyclonality with respect to these oncogenes is a phenomenon limited to a few PTC cases. However, it is dutiful to highlight that FISH and PCR-based analysis are concordant assigning to ITH a significant impact in PTC, while NGS analysis are discordant. More recently, we analyzed a large cohort of 208 PTC by MassARRAY, and we calculated the allelic frequencies of BRAF V600E and RAS mutations by subtracting non-tumor cells. The majority of cases had an allelic frequency of the mutated allele consistent with a monoclonal origin of the tumor, suggesting the occurrence of ITH in a small, though not negligible, subset of tumors (8%) [20].

Spatial Heterogeneity
Impact on clinical behavior: some studies showed that mutation density highly correlates with aggressive histologic features and risk of recurrence [20,22]. Although the association of mutation density with worst outcome was not found in one study including cases with high-recurrence risk [71], most studies report that patients with dual mutations are associated with an older age at diagnosis and a worst outcome, suggesting that tumors with multiple mutations undergo a positive selection and are more aggressive [40,[43][44][45]47]. Of note, TERT mutations were found to be mainly subclonal in PTCs, whereas they were clonal in poorly differentiated and anaplastic tumors, consistent with a positive selection during tumor evolution [54]. As far as BRAF V600E concerns, subclonal mutations were associated with smaller PTC tumors [29,37,53,57,72], lower extrathyroidal extension [37,53] and lower recurrence rate [57], while other studies did not report significant association with disease progression [7]. In contrast, Masoodi et al. showed that subclonal mutations were significantly higher in relapsed PTCs and that cases with high burden of subclonal mutations were associated with distant metastasis and increased risk of relapse or death [58].
Impact on diagnosis: as stated for other tumors [73], the presence of molecular heterogeneity highlights the importance of sampling multiple areas of the same tumor to better ascertain the range of genomic alterations characterizing its progression. In a recent study, it was found that the 23.1% of the somatic mutations found in a large PTC series were not identified in all regions of the tumor, revealing that ITH limits the diagnostic value of single diagnostic biopsy sample [58]. Particularly in advanced TCs, the heterogeneous presence of genomic alterations may impair patient genotyping and subsequent prognostic classification and targeted therapy. Moreover, some mutations, despite their very low allelic frequencies in the primary tumor, can be responsible for the development of metastases, indicating the need for highly sensitive diagnostic tools to obtain a full genetic characterization.
Impact on treatment: ITH could have a significant impact on prognosis and treatment response and may influence the best and targeted therapeutic strategy, especially in the era of personalized medicine [5]. Indeed, the genetic pattern found in the primary tumor that in some cases directed the clinical and therapeutic decisions, may be not representative of the variation within a tumor as a whole, and may evolve during tumor progression, also due to the selection pressure of treatment [4]. Consequently, for many tumors, combinations of targets-based drugs will likely be necessary to control the tumor growth.

Temporal Heterogeneity
The different subclones are predicted to have different resistance mechanisms to treatment. If they are all sensitive to the initial treatment, the tumor will be eradicated. Cancer relapse can be due to the progressive enrichment with time of drug resistant cancer cells already present in the heterogeneous cancer cell population. Indeed, it is predicted that a fraction of cells with stem cell properties and probably also a fraction of adult cells within the heterogeneous population are drug resistant, able to survive to treatment and expanding over time [74], as shown for some cancers such as lung [75]. It is worth mentioning that temporal heterogeneity derives also from the progressive increase of the mutational burden during the de-differentiation process [15][16][17][18]. The best way to identify the presence of different subclones, is to submit the primary tumor to an ultra-deep sequence which allows identifying all the mutated clones, in order to start a treatment directed by the characteristics of the dominant clone as well as the rare resistance clones, with a combination of therapies to eradicate all clones. Nevertheless, this option is rarely applied, especially for TC which is a tumor well curable in the vast majority of cases.
Interesting insights come from a recent report of an acquired KRAS mutation that developed during treatment with BRAF and MEK inhibition in a patient with a BRAF-mutated PTC. The KRAS mutation was detected at the time of progression in the peripheral blood, too [76]. This last finding demonstrates that resistant mutations can be documented even with non-invasive methods, such as plasma DNA or circulating tumor cells [77].
Additional tools to identify resistant subclones, imply their analysis in cellular systems, which also allows the testing for different and possible novel therapeutic options. As far as TC concerns, Antonello et al. expanded a sub-population of cells with primary resistance to vemurafenib and found that they harbor amplification of chromosome 5 and mutations in RBM genes which are crucial for genome stability during cell division [78]. A combined therapeutic approach using BRAF V600E and CDK4/6 inhibitors was able to induce apoptosis in both naïve and vemurafenib-resistant cells, indicating that this combined therapy could be tested in a clinical trial of advanced TC patients.

Conclusions
Intratumoral genetic heterogeneity identify a phenomenon by which a neoplasm harbors genetically different subclones that may be intermixed or spatially separated. Importantly, the number of genomic alterations spontaneously increases with tumor progression and can evolve in response to treatments. Different types of cancer are characterized by different subclonal complexity, sometimes with a high number of subclones such as melanoma, and other times showing a low number such as the case of thyroid cancer. The existence of the ITH is well accepted for advanced PTC, while it is a debated issue for PTCs in the early stage of progression. Several data presented in this review indicate that ITH is not a marginal event that can occur in PTC at any step of development, including early stage PTC. Furthermore, the demonstration of ITH of BRAF and RAS mutations and RET/PTC rearrangements highlights the existence of unknown genetic or epigenetic alterations that still need to be identified.
The ITH of BRAF and RAS mutants and RET/PTC rearrangements in early stage PTC, call into question their role of driver mutations and initiators of thyroid carcinogenesis. In a stochastic model of multi-step carcinogenesis, the coexistence of BRAF mutation negative and positive subclones entails that these oncogenes are generated succeeding other unidentified genetic alterations which have the role of initiators. In the cancer stem cell (CSC) model, a small subpopulation of CSCs that can self-renew and differentiate to produce phenotypically diverse cancer cells acquires initiating mutations, with BRAF, RAS and RET/PTC alterations developing later, and involving only a cellular subpopulation.
Based on the present knowledge of this topic, the definition of the TC genetic background must consider the existence of either heterogeneity or multiple mutations with different allelic percentage. In this context, benefits will come from NGS techniques (especially whole genome sequencing) since these tools allow better appreciating clonal cancer cell fractions of important mutations mostly because of the high-resolution CNV analyses. It is also well known that the presence of intra-tumor heterogeneity is the cause for relapsing after the therapy, which effectively eliminated the competition of the major clone, leaving minor subclones free to expand. The assessment of such complexity, for instance by the analysis of paired primary and metastatic samples from the same patient to acquire insights into clonality and subclonality patterns of genomic events, is mandatory for the further development of personalized medicine in TC.
Since novel generation compounds highly selective for a specific genetic alteration are currently on trial, the assessment of the genetic pattern of malignant tumors, including advanced TC, is definitely needed. The genetic evaluation should consider the identification of the percentage of mutated cells, too, by normalization for the non-tumoral cell content of the neoplastic mass, since it is predictable that different allelic frequencies of a given mutation could correlate with the response to treatment.

Conflicts of Interest:
The authors declare no conflict of interest.