Real-Time Genomic Characterization Utilizing Circulating Cell-Free DNA in Patients with Anaplastic Thyroid Carcinoma

Introduction

Anaplastic thyroid carcinoma (ATC) is the most aggressive tumor of the head and neck region (1 –3). Its biological behavior diverges sharply from that of well-differentiated thyroid cancers, which are generally curable with surgery alone and which are associated with excellent disease-free and overall survival (4). Because of its rapid growth and aggressive pattern of regional and distant metastasis, ATC generally presents at an advanced stage. Multimodality treatment (i.e., surgery combined with chemoradiotherapy) is essential for achieving locoregional control and occasionally a cure (1 –3). Unfortunately, too often, patients newly diagnosed with ATC are deemed incurable and slated for palliative treatment or hospice (5).

Over the last decade, an enhanced understanding has been gained of the genomic and molecular background of ATC (4). Although full exome sequencing of an increasing number of ATC tumors is currently underway as part of at least two multi-institutional collaborative groups, there is a reasonable understanding of the genomic signature of ATC (6 –8). This has allowed the generation of several novel therapeutic approaches currently under investigation at the authors' institution and other tertiary centers. As these prospective clinical trials continue to expand, a debate has arisen about the optimal method for profiling ATC tumors prior to initiation of treatment.

At the University of Texas MD Anderson Cancer Center (UTMDACC), patients with a new diagnosis of ATC undergo tumor characterization termed the CM50 assay that uses next-generation sequencing (NGS) to screen for known somatic mutations in a panel of 50 commonly mutated genes as part of the standard of care. In addition, patients also undergo evaluation for circulating cell-free DNA (cfDNA) using a commercially available kit (9). This dual interrogation of the tumor genome is performed for several reasons. First, tumor retrieval from outside institutions and mutation testing with CM50 can take several weeks. Therefore, treatment decisions can be delayed if CM50 alone is relied upon. Second, solid tumors are often heterogeneous with respect to clonal expansion. Significant differences in primary and metastatic tumor genetic profiles can be found, therefore requiring a needle biopsy and genetic testing of the distant metastatic disease, which can have inherent risks. Third, large solid tumors contain significant fractions of necrotic, nonviable, or borderline viable tissue with heavy inflammatory infiltrate. These phenomena lead to lower sensitivity for mutational detection from fine-needle aspirations and core-needle biopsies in ATC compared with other solid tumors (10 –13). Clonal heterogeneity can confound even data generated from surgical specimens. Unlike most other solid tumors, ATC can occur de novo or through de-differentiation of more indolent cancers such as papillary thyroid carcinoma (PTC) and follicular thyroid carcinomas (FTC) (4,14). ATC tumors can still maintain areas of differentiation consistent with the tumor of origin. As a result, sampling error represents a significant diagnostic/genetic profiling challenge in this disease.

The introduction of cfDNA evaluation provides clinicians and researchers with the unique ability of sampling tumor DNA that is shed into circulation in order to obtain real-time diagnostic or therapeutic targeting information (15,16). Utilization of cfDNA can potentially supplement interrogation of tumor specimens by providing a potentially more comprehensive picture of the genomic landscape for each individual ATC patient. This retrospective study sought to evaluate data from patients with a diagnosis of ATC and to compare results from tumor-based sequencing platforms and a commercially available blood-based cfDNA platform. This study is meant to provide pilot data regarding the potential clinical relevance of this dual interrogation method in the development and implementation of novel ATC treatment strategies.

Methods

Patients included in this analysis were evaluated clinically at the UTMDACC between August 2015 and April 2016. All patients carried a diagnosis of ATC, which was confirmed by trained head and neck pathologists at the UTMDACC. Genomic analysis was performed using previously available surgical specimens or tissue obtained at the time of diagnosis via fine-needle aspiration, core biopsy, or surgical biopsy using a UTMDACC NGS platform (CM50). In one case (patient #12), an outside report was utilized to evaluate the tumor genomic profile (Foundation One) (17). CM50 is a NGS-based platform in the CLIA-certified molecular diagnostics laboratory at UTMDACC that utilizes the Ion AmpliSeq™ Cancer Hotspot Panel (Life Technologies, Carlsbad, CA) for detecting point mutation, short insertions and deletion, and high level of amplification in the coding sequence of a total of 50 genes from tumor tissue (Supplementary Table S1; Supplementary Data are available online at www.liebertpub.com/thy) (18,19). As normal tissue is not utilized in this assay, the ratio of allelic changes and reference to the dbSNP polymorphism repository aid in noting variants unlikely to be tumor associated. A minimum coverage depth of 250 reads per amplicon is required with a variant frequency of 5% and higher with a minimum of 25 reads with the variant is considered positive. The variants detected by this platform are classified into four groups based on both analytic findings such as allelic frequency and the currently available information in publicly available and periodically curated reference databases COSMIC v64 (Catalog of Somatic Mutations in Cancer, Wellcome Trust Sanger Institute, Cambridge, United Kingdom) and dbSNP v137 (National Institute of Health, Bethesda, MD).

Patients underwent a blood draw as part of their treatment evaluation. Testing of patient blood was performed using a commercially available cfDNA platform (Guardant 360 [G360]) as was previously described in detail by Lanman et al. (9). Briefly, whole blood was collected in two Streck cell-free DNA BCT^® tubes and shipped overnight at ambient temperature to the Guardian CLIA-certified laboratory facility. cfDNA was extracted from plasma, and genomic alterations were analyzed by a proprietary digital sequencing technology by parallel paired end synthesis-by-sequencing (8000× average coverage depth) of amplified target genes utilizing an Illumina Hi-Seq 2500 platform complemented by systematic end-to-end process optimization. Analysis focused on complete exon sequencing of 29 genes and critical exon coverage for an additional 39 genes along with copy number amplification in 16 genes and fusions involving ALK, RET, ROS1, and NTRK1, and EGFR insertion/deletion mutations (Supplementary Table S2) (9).

All testing and data collection was performed as part of standard-of-care treatment, but the analysis of the data was performed under an Institutional Review Board–approved protocol in compliance with institutional guidelines. Concordance between the two platforms was calculated as follows: concordance = # of concordant mutations/(# of concordant mutations + # of discordant mutations).

Results

Patient characteristics are summarized in Table 1. Twenty-two of these patients were new to the institution. One was a re-referral for suspected recurrence. Fourteen patients had undergone one surgery for ATC prior to presentation, most commonly a total thyroidectomy. Prior to presentation at the UTMDACC, four patients underwent multiple surgeries, three underwent radioactive iodine treatment, six underwent external beam radiation therapy (EBRT), and seven underwent chemotherapy treatment. One patient (#10) presented with multiple primary tumors, including a concomitant esophageal primary that required chemotherapy and surgery prior to presentation. Twenty patients had active regional disease at the time of presentation, 11 of whom also presented with distant disease (Table 1). At the time of last contact, 15 patients were alive, and eight patients had died. Mean and median follow-up times for the cohort were 142 and 102 days, respectively. All patients underwent evaluation using both NGS platforms. Time from request to completion of NGS report was significantly shorter for the G360 platform (M = 13 days, median = 13 days) compared with the CM50 platform (M = 19 days, median = 18 days; p = 0.009).

Treatment decisions for all patients are summarized in Table 1. One patient (#2) was disease free at time of evaluation based on anatomic imaging and was slated for observation. Three patients opted for hospice. The remaining patients all received treatment subsequent to presentation and cfDNA analysis. Some patients underwent sequential treatment using different approaches/agents (patients #4, 15, 20, 21, and 23).

The two NGS platforms did not demonstrate complete gene and codon overlap (Supplementary Tables S1–S4). Genes represented on both platforms that demonstrated mutations in at least one patient are summarized in Table 2. For the purpose of analysis, the patients were divided into three groups. Group 1 included patients with distant and/or locoregional disease without treatment prior to dual platform NGS analysis. Group 2 included patients treated with surgery for locoregional disease and minimal residual disease or who received recent cytotoxic chemotherapy for distant metastatic disease at the time of dual platform NGS analysis. Group 3 included patients without evidence of active disease at the time of dual platform NGS analysis. As expected, patients in group 3 demonstrated the lowest concordance between the two platforms, since the patients did not have detectable disease at the time of cfDNA analysis. Patients in group 1 demonstrated the highest concordance (BRAF 7/7, TP53 4/6, PIK3CA 3/5, NRAS 3/3) between the two platforms (0.72). Patients in groups 2 and 3 demonstrated much lower concordance (0.06 and 0.00, respectively). It is important to note that for the purposes of this study, the official calls generated by the NGS platforms were used in accordance with their current clinical standard for reporting. Secondary analysis of the CMS50 raw reads did identify concordant mutations present below the clinical detection threshold in three additional patients that were also found in the cfDNA analysis. These data are being used internally to develop more inclusive NGS analysis of tumor tissue and to optimize future clinical sequencing efforts, but are outside of the scope of this article.

OPEN IN VIEWER

Table 2.

Tumor Profile

Genomic alterations in the 23 patients are shown above for genes that were covered on both NGS platforms. Group 1 includes patients with distant and/or locoregional disease without treatment prior to dual platform NGS analysis. Group 2 includes patients treated for locoregional or distant disease, with minimal residual disease at the time of dual platform NGS analysis. Group 3 includes patients without active disease at the time of dual platform NGS analysis.

Tu, NGS analysis of tumor; Pl, NGS analysis of plasma (cfDNA); CNC, codon not covered; NGS, next-generation sequencing.

The most commonly mutated genes were TP53 in 15/23 (65%) and BRAF in 11/23 (48%), consistent with previous reports (6,20,21). In general, the two platforms detected the same TP53 mutation. The exceptions may reflect the fact that although some mutations are present in the primary tumor, they do not generate a sufficient cfDNA signature to allow for detection. Conversely, since the cfDNA is very sensitive, it may identify clones that are at a low enough frequency that they escape the detection threshold of the other platform. The distribution of TP53 mutations is similar to that described in other solid tumors, and only the R175H mutation was identified in more than one patient. Patients #5 and #13 demonstrated discordant TP53 mutations, and patients #5, 6, 9, and 13 demonstrated multiple TP53 mutations identified between the two NGS platforms (Supplementary Tables S3 and S4).

BRAF alterations consisted primarily of the V600E mutation (9/11), with the exception of patients #20 and #22. In addition to BRAF mutations, four patients demonstrated alterations in NRAS, one an alteration in KRAS, and one an alteration in HRAS. BRAF mutations did not overlap with alterations in HRAS, KRAS, or NRAS. One patient (#23) demonstrated alterations in both KRAS and NRAS. PIK3CA and PTEN mutations were detected on both platforms. Half of the patients with PIK3CA mutations also demonstrated alterations in BRAF, and 4/6 patients also demonstrated alterations in TP53. One patient had no detectable alterations in either platform.

Discussion

Improvements in clinical outcomes for ATC and other aggressive malignancies require optimization of existing multimodality regimens and development of new treatment strategies (2). A solid understanding of the genomic profile of these tumors is essential in order to achieve both clinical goals. In ideal circumstances, all patients would be able to undergo a complete molecular profiling of their tumor at the time of initial presentation, including a comprehensive evaluation of genomic and epigenetic events. However, this is not feasible in the clinical setting for several reasons. First, a significant number of patients with ATC are not definitively diagnosed until the tumor is removed and a complete pathologic examination is completed. If the tumor is not appropriately stored or is not available for sequencing at a later date, sequencing information for these patients may not be available to the responsible clinicians, especially when decisions about adjuvant treatment (e.g., chemotherapy, radiation) are made. This was the case for the majority of the patients evaluated in this study. In these cases, decisions regarding additional treatment can now be made using information derived from the primary tumor, as well as from residual or recurrent disease that requires further treatment. Second, sampling of ATC tumors prior to surgical intervention can target nonviable or heterogeneous tumor tissue and therefore fail to be informative. Third, comprehensive sequencing and epigenetic analysis remains prohibitive not only in terms of cost, but also with respect to integrated data analysis and timely clinical decision making. As a result, clinicians must focus on utilizing available technology in order to inform treatment decisions best at the time of patient presentation.

This study found that patients who were treatment naïve (group 1) had the highest concordance between tumor and blood-based platforms. Notably, there was 100% concordance between the two platforms for the presence of a BRAF^V600E mutation in this group of patients. Patients who underwent surgical resection with minimal residual disease or cytotoxic chemotherapy (group 2) at the time of cfDNA testing had CM50 testing performed on tumor that was sampled before treatment for ATC. In the case of the patients who underwent surgery, this was likely due to the driver mutation not being present in high enough levels to detect in cfDNA. In these patients, cfDNA is less useful for clinical decision making. However, in patients who underwent cfDNA testing after cytotoxic chemotherapy, it is possible that discordant results occurred due to subclone emergence as a result of drug resistance. In this scenario, cfDNA can potentially be more representative of the post-treatment tumor mutations. Because post-chemotherapy specimens were not available for analysis, it is not known if the concordance would have been higher between the two platforms in this patient population. Another important finding was the statistically significant faster turnaround for the cfDNA NGS platform. Physicians treating this devastatingly aggressive cancer realize that a difference of a few days to obtain mutation results can be an advantage.

Two novel therapeutic approaches are increasingly available to patients with ATC through tertiary referral centers. The first is a targeted approach utilizing molecularly targeted therapy based on mutation data. A good example of the success of this approach in ATC is the utilization of BRAF inhibitors in BRAF^V600E -mutated tumors. Several trials with BRAF inhibitors (NCT 01524978 and NCT 02091141) and the combination of BRAF inhibitors with MEK inhibitors (NCT02034110) are in various stages of enrollment. The results of one of these trials have already been reported in the literature (22). Of note, five of the patients who underwent sequencing in the present series were treated with BRAF-directed therapies. Other examples of clinical trials that are recruiting ATC patients based on the presence of a genetic abnormality are investigating the efficacy of drugs targeted against HRAS (NCT02383927) and ALK (NCT02289144). In the present patient cohort, PIK3CA alterations were often noted in the context of BRAF or RAS alterations. These data may be utilized in subsequent trial design of be taken into consideration for patients who fail clinical trials consisting of BRAF and MEK inhibition.

A second approach to the treatment of patients with ATC is a microenvironment-targeted approach, for example with VEGFR inhibitors such as lenvatinib (NCT 02657369), especially in patients without targetable mutations (23). To that end, both tumor and the cfDNA platforms can provide useful information for treatment selection. Specifically, patients without identified actionable mutations may be slated for microenvironment-targeted therapies or maximal conventional standard-of-care treatment. Although this comparative effectiveness is speculative at this time, sequencing information may still provide useful information in patient selection for these and other future clinical trials. It is recognized that whole genome sequencing may have identified additional actionable mutations that may have been missed using the clinical platforms described in this study. However, these two current platforms provide a relatively comprehensive analysis of those genes commonly implicated in cancer development, and both remain under continued development and refinement.

EBRT is an essential component of multimodality treatment aimed at achieving locoregional control in ATC (1,2). Mutation of the TP53 gene has been associated with radioresistance in multiple solid tumor histologies, including studies by the authors' group in head and neck squamous-cell carcinomas (24 –26). Given the frequency of TP53 mutations identified in this cohort, as well as by other groups, it is paramount to clarify the oncogenic role of TP53 mutations in ATC development and to develop novel radiosensitizing strategies aimed at overcoming TP53 resistance. One particular interesting aspect of the TP53 data summarized above is the partial discordance identified with respect to the specific TP53 mutations between the two platforms. This may reflect a broader heterogeneity in TP53 mutations across ATC tumors or development of multiple mutations in the same tumor population as described in some solid tumor cell lines (27,28). Clarification of which mechanism is primarily responsible for the observed clinical phenotype is the subject of ongoing basic science studies in the authors' group.

Although TP53 and BRAF are by far the two most commonly mutated genes in ATC, both panels detected other potential targets such as PIK3CA, HRAS, and PTEN. This information may be useful for the development and implementation of Phase I/II trials evaluating novel compounds across multiple rare histologies. Interestingly, BRCA1/2 mutations were identified initially on the cfDNA panel in multiple patients, which, to the authors' knowledge, has not been previously reported. The significance of these mutations is unclear at this time and will require additional investigation. The next generation of the cancer mutation panel at the authors' institution will include BRCA1/2 among the analyzed genes, and is expected to provide more data regarding the frequency of this genomic alteration in ATC tumors.

The present study has several limitations. First, this is a limited patient cohort, which is not unusual given the rarity of the disease process. Nevertheless, this represents a significant increase in the institutional volume, considering the entire cohort was evaluated in approximately eight months. Second, the timing of the CM50 and G360 testing is discordant in several patients with respect to when the tissue tested was obtained (primary tumor vs. recurrent/residual disease). Third, although both sequencing platforms are validated for clinical testing, additional validation testing is not available at this time.

Based on the data from this limited patient cohort, it can be concluded that genomic testing using available sequencing platforms is useful for clinical decision making for ATC patients being considered for targeted therapies. Larger data sets generated in the setting of these trials will then also provide essential information for novel drug design as a more comprehensive picture of the mutational landscape of ATC is obtained. Although occasionally discordant, utilization of both platforms has the ability to generate the maximum amount of information available for optimal trial selection. The significance of detection of rare clones noted on cfDNA (<0.3%) testing is unknown at this time, but may also provide information regarding potential treatment resistance and tumor recurrence. Following continued validation of this approach, cfDNA testing may be considered the primary method of characterizing ATC genomic information due to its rapid turnaround, increasingly widespread clinical availability, and decreased requirement for invasive tissue sampling. It should be noted that recent changes in G360 testing policy are expected to reduce turnaround time by approximately 25%, and may result in a significant improvement in time to report compared with the current tissue-based platform. It is recognized that increasing diagnostic expenditures may not be sustainable in routine clinical practice in the absence of a proven clinical benefit. As such, the recommendation for molecular testing is primarily limited to the clinical trial setting where it may assist in treatment selection and optimization of future trial design.