Preoperative detection of the TERT promoter mutations in papillary thyroid carcinomas

Telomerase reverse transcriptase promoter (TERT‐p) mutations are strongly associated with tumour aggressiveness and worse prognosis in papillary thyroid carcinomas (PTCs). Since the TERT‐p mutations have been reported to be subclonal, it is unclear how accurately they can be detected by preoperative fine‐needle aspiration (FNA). The objective of this study was to analyse the concordance rate of the TERT‐p mutations between preoperative FNA and corresponding postoperative surgical specimens.


| INTRODUCTION
The incidence of thyroid cancers has increased in recent years.
This increase is mostly attributed to papillary thyroid carcinomas (PTCs). 1 Although PTCs usually have a favourable prognosis, 10%-15% of patients have recurrences, some of which become refractory to treatments. 2,3 The BRAF V600E mutation is the most common genetic alteration in PTCs. Its prevalence varies from 30% to 80%, 4 which seems to be dependent on the study population. Many studies have reported an association between the presence of the BRAF V600E mutation and aggressive clinicopathological features in PTCs. [5][6][7] On the other hand, there are also studies, especially from Japan, demonstrating no such relationship. 8,9 Thus, the clinicopathological significance of BRAF V600E is not universal and still remains controversial.
Two somatic mutations, located −124 and −146 base pairs upstream from the ATG start site (−124C>T and −146C>T) of the telomerase reverse transcriptase (TERT) gene have been found in PTCs. 10,11 They are mutually exclusive, and the average frequency of either mutation in PTCs has been reported to be about 10%, 12,13 which seems to have no major difference between populations. The TERT promoter (TERT-p) mutations strongly correlate with older patient age. 9,[14][15][16][17][18][19][20] Many studies have consistently demonstrated that the coexistence of the BRAF V600E mutation and the TERT-p mutation is strongly associated with aggressive clinicopathological characteristics and a worse prognosis. 9,15,17,[21][22][23][24][25][26][27] So far, the TERT-p mutations are the best prognostic molecular marker in PTCs. Therefore, their preoperative detection in fine-needle aspiration (FNA) specimens could be useful in the determination of PTC management.
The TERT-p mutations have been reported to be an early event during cancer progression in glioblastoma, hepatocellular carcinoma and melanoma, whereas in thyroid cancer, they have been reported to be a late event. 28,29 If the TERT-p mutations are a late event, only a fraction of cancer cells harbour the mutation, and the frequency of the mutant allele is supposed to be low.
Indeed, according to The Cancer Genome Atlas (TCGA) data, the allele frequencies of the TERT-p mutations were 5%-50% (23% on average), 30 suggesting that the TERT-p mutations are not a clonal event in many PTCs. Based on these findings, it is unclear whether preoperative FNA can be effectively used to detect the TERT-p mutations.
There are four studies that preoperatively analysed the TERTp mutations using FNA or core-needle biopsy (CNB) samples in thyroid carcinomas. 23,[31][32][33] These studies evaluated the TERT-p mutations in only preoperative FNA or CNB specimens, but comparison with postoperative surgical tissues was not performed. Considering that the TERT-p mutations are not always clonal in PTCs, it is not clear how accurately the TERT-p mutations could be detected in preoperative FNA samples. In this study, therefore, we analysed the concordance rate of the TERT-p mutations between FNA and surgical specimens. The study protocol was approved by the institutional review boards of Nagasaki University and Kuma Hospital. Written informed consent was obtained from each patient. All methods were carried out in accordance with the relative guidelines and regulations.
Ultrasound-guided FNA was performed by a well-experienced operator using a 22-gauge needle fitted to a 10-ml syringe. Most of the material from the needle was used for cytological examination, and the remaining was used for DNA extraction by washing needles with 1 ml of RNA later (Thermo Fisher Scientific). All samples were stored at 4°C for up to 1 month. DNA was extracted using ISOGEN reagent (Nippon Gene). The corresponding postoperative formalinfixed and paraffin-embedded (FFPE) specimens were obtained, and DNA was extracted from manually microdissected tumour portions using a QIAamp DNA Mini Kit (Qiagen). For each case, we used 1-4 serial sections of 10-µm thickness, depending on the tumour size.

| Immunohistochemistry
Manual immunohistochemistry was performed on 4-μm-thick sections of FFPE tissues using mouse monoclonal anti-BRAF (mutated V600E) antibody (clone VE1; Abcam). The sections were incubated with the VE1 antibody at a 1:100 dilution. Antigen retrieval and detection were performed using cell conditioning solution (Roche Diagnostics) and Novolink Polymer Detection System (Leica). NAKAO ET AL.

| Statistical analysis
QuantaSoft software (Bio-Rad Laboratories) was used to determine the copy number data and other relevant information from ddPCR runs. The Fisher's exact test or Fisher-Freeman-Halton exact test was used for categorical data, and the Mann-Whitney test was used for continuous data to compare between any two (sub)groups. Multivariable linear, logistic or ordinal response regression models were used to analyse the associations between the TERT-p status and clinicopathological parameters. Logistic regression models with very small numbers of outcomes (<5 per cell) were conducted using Firth's approach to bias-reducing penalised maximum likelihood fit using the LOGISTIC procedure in the 9.4M5 version of SAS (SAS Institute).
Multivariable linear and ordinal response models were run using the GENMOD procedure using appropriate distributions and link functions. All p-values were two-sided and considered significant if p < .05.

| Establishment of the ddPCR assay
Our ddPCR assay specifically detected the wild-type and mutant alleles of the TERT-p and BRAF (File S1). To evaluate the quantitative performance of the assay, serially diluted mutant PCR products were mixed in wild-type PCR products to obtain various reference samples with mutation fractions of 0%, 1%, 5%, 10%, 50% and 100%. For each, four different sets of total DNA copies F I G U R E 1 Detection of the TERT-p mutations in FNA and FFPE samples. The TERT-p mutational status was determined based on the number of total TERT-p DNA copies in a reaction (horizontal axis) and the percentage of the TERT-p mutation allele (vertical axis) using pre-established cutoff limits (Supporting Information). Each sample is represented by an open/closed triangle/square depending on the concordance with the corresponding counterpart. (A) FNA, (B) FFPE results. FFPE, formalin-fixed and paraffin-embedded; FNA, fine-needle aspiration; TERT-p, telomerase reverse transcriptase promoter; wt, wild-type of 500, 1000, 3000 and 9000 were prepared and analysed. The expected and observed mut% were well correlated in all tested samples of both genes, and all coefficients of determination were greater than 0.989 ( Figure S1).
Since the TERT-p mutations are C>T transitions, artificial C>T mutations during PCR caused by hydrolytic deamination of cytosine are a problem in precise detection of the TERT-p mutations, especially in stored DNAs. The cytosine deamination creates uracil, and uracil can pair with adenine during elongation, leading to the C>T mutation and causing a false positive. We used uracil DNA glycosylase to avoid the false positive; however, it was not possible to completely eliminate it. Therefore, the cutoff limit for detection of the TERT-p mutation in the ddPCR assay was statistically determined by analysing serially diluted mutant PCR products mixed with wild-type PCR products and 100% wild-type PCR products. We determined mut% that yielded the number of mutant signals significantly exceeding the background, calculated the 95% prediction interval for those values, and used its upper boundary as the cutoff; all samples displaying TERT-p mut% above the cutoff were considered TERT-p mutationpositive. See File S2 with Figure S2 for detailed calculations.
In addition, we determined the minimal DNA copy number in a reaction that was necessary to correctly distinguish the true negative samples from those showing no TERT-p mutant signals due to an insufficient amount of template DNA. Under the conditions used in our ddPCR, the minimal number of TERT-p DNA copies in a reaction was more than 64 (File S2). Samples with the total TERT-p DNA copy number ≤64 were considered 'inadequate'.

| Analysis of FNA and FFPE samples by the ddPCR
We recruited patients equal to or older than 55 years old to perform an efficient comparison as the presence of the TERT-p mutation is strongly associated with age. This age is also used as a threshold in the American Joint Committee on Cancer (AJCC) eighth edition staging system for risk stratification. A total of 96 nodules with pathologically confirmed PTC were analysed. Based on the cutoff limits, 19 (20%) and 4 (4%) of the FNA samples were found to carry the −124C>T and −146C>T mutations, respectively, 66 (69%) were wildtype and 7 (7%) were inadequate ( Figure 1A). Of the corresponding 96 FFPE samples, 24 (25%) and 4 (4%) had the −124C>T and −146C>T mutations, respectively, and 68 (71%) were wild-type ( Figure 1B). All FFPE samples were successfully genotyped. Assuming  (Table 1). Cohen's kappa was 0.834 (p = 2.52E−15), which corresponds to 'almost perfect agreement'. 34 A total of six cases had discordant results: one case was FNA mut/FFPE wild-type and five cases were FNA wild-type/FFPE mut.
The BRAF V600E mutation was also examined in the same fashion.
The BRAF V600E mutation is a T>A transversion, and we did not observe any false positive in our ddPCR assay. Of the 96 FNA samples, 85 (89%) were found to carry the BRAF V600E mutation and 7 (7%) were wild-type. There were four (4%) inadequate cases. Of the 96 corresponding FFPE samples, 91 (95%) had the BRAF V600E mutation and 5 (5%) wild-type. All FFPE samples were successfully genotyped.

| Relationship between the mutational status and clinicopathological features
Initially, it is important to note that all cases were ≥55 years old (high age group in the AJCC eighth edition staging system) and all TERT-p mutation-positive tumours also had BRAF V600E mutation (so-called  Correspondence analysis demonstrated that the mutational status of TERT-p and BRAF determined by FNA and that by FFPE were highly correlated. Other parameters showing a high correlation with the mutational status were higher Ki-67 labelling index and age more than 69 years old ( Figure S3).

| The clonality of the TERT promoter mutation
The allelic frequency of the TERT-p mutation does not accurately tell us the clonality of the TERT-p mutation in cancer cells because tumour tissue contains not only tumour cells but also stromal, endothelial and blood cells. Then, we used the allelic frequency of the  Figure S4).
We calculated the proportion of the TERT-p mutation-positive cancer cells (TP) by dividing the allelic frequency of the TERT-p mutation by that of the BRAF V600E mutation. In the 28 FFPE samples that harboured the TERT-p mutation, the TPs varied from 51% to 136% ( Figure 2), although in most cases they were 80%-120%, suggesting that the TERT-p mutation may be nearly clonal in most PTCs. Two cases had higher TPs (135.0% and 135.5%), and three cases had lower TPs (51.1%, 52.8% and 65.7%), which could be due to copy gain or loss of heterozygosity and low clonality, respectively. We also analysed the association of TP with clinicopathological characteristics. As shown in Table 4, the higher TP was significantly associated with the higher pT category of both the seventh (multivariate p =.023) and eighth (multivariate p = .005) editions of AJCC TNM classification and extrathyroidal extension (univariate p = .015 and multivariate p =.05).

| DISCUSSION
In this study, we analysed the concordance rate of the TERT-p mutations between preoperative FNA specimens and their postoperative counterparts using a highly sensitive ddPCR technique.
There have been three studies that preoperatively analysed the TERT-p mutations using FNA samples in thyroid carcinomas. 23,31,32 According to these reports, the frequency of the mutations ranged from 4.5% to 14.5%, but none of these studies assessed the performance of analysing the mutations using FNA by comparing them with surgical specimens. Our present study is the first report to evaluate the concordance rate of the TERT-p mutations between preoperative FNA specimens and their postoperative counterparts.
In the present study, the frequencies of the TERT-p mutations in FNA and FFPE samples were 24% and 29%, respectively, which was higher than in previous reports. One of the reasons is that we recruited PTC patients aged 55 years or older since the TERT-p mutations have been reported to correlate with older patient age. [14][15][16][17][18][19] Another reason for the higher frequency of the TERT-p mutations may be the use of ddPCR assay, which is more sensitive than conventional Sanger sequencing and real-time PCR.
The TERT-p mutations were significantly associated with age, tumour size, extrathyroidal extension and the Ki-67 labelling index in not only the results obtained from the FFPE sections but also those from the FNA samples. Therefore, the mutational status of the TERT-p analysed using preoperative FNA samples can be used This analysis may enable us to select an optimal strategy, such as applying total thyroidectomy from the first operation even with a small PTC with no nodal metastasis.
The sensitivity of the TERT-p mutations in FNA specimens was 81.5%. We had expected that the detection rate of the TERT-p mutations in FNA specimens could have been lower because the TCGA data showed that the mutant allele frequencies of the TERT-p mutations were 5%-50% (23% on average). 30 They used driver oncogenes to calculate tumour purity to exclude normal alleles from nontumour cells, which is basically the same method like ours. In this study, we calculated the TP using the allelic frequency of the BRAF V600E mutation to adjust for tumour purity. The mean TPs were 80%-120% in most of the cases, suggesting that the TERT-p mutations are a nearly clonal event in many PTCs. We do not have a good explanation for this discrepancy between the TCGA data and our present results. We had two cases with higher TP (135.0% and 135.5%), and gene amplification may be involved in these cases. On the other hand, there were three cases with lower TP (51.1%, 52.8% and 65.7%), and in these cases, the TERT-p mutations were considered to be subclonal. We also found a significant association of the TP value with higher pT and extrathyroidal extension, suggesting that the TP may be correlated with aggressive properties.
In some cases, the mutational status was not consistent between FNA and FFPE samples. All five FNA specimens that were FNA wildtype/FFPE mut for the TERT-p mutation had sufficient amounts of DNA (see Figure 1A), and their TPs were not low (78.8%, 84.0%, 86.7%, 94.9% and 135%). For the BRAF V600E mutation, there were three FNA wild-type/FFPE mut cases. One possible explanation for the results is the absence of mutation-positive cells in the solution obtained by washing needles. If so, it might be possible to increase the sensitivity by using the whole contents in the needle. The frequency of the TERT-p mutation in one FNA specimen (FNA mut/ FFPE wild-type) was just above the cutoff value. Cases that are close to the cutoff value should be judged carefully, and, if possible, retesting is recommended.
There are some limitations of this study. First, the confusion matrix was composed assuming that the results of the TERT-p mutations in the FFPE samples were true. However, one can imagine that the FFPE sections may not represent a whole tumour.
Second, due to the short observation period, we were not able to examine the association between the mutational status and the risk of recurrence.
In summary, the mutational status of the TERT-p in the FNA samples detected by our ddPCR assay was highly concordant with that in the postoperative FFPE sections. Our test method had high sensitivity and specificity, and the results were significantly associated with conventional clinicopathological parameters correlated with disease aggressiveness. Therefore, preoperative detection of the TERT-p mutations in FNA specimens may provide an option for selecting the appropriate surgical procedure.