Classification of follicular-patterned thyroid lesions using a minimal set of epigenetic biomarkers

OBJECTIVE

The minimally invasive fine-needle aspiration cytology (FNAC) is the current gold standard for the diagnosis of thyroid nodule malignancy. However, the correct discrimination of follicular neoplasia often requires more invasive diagnostic techniques. The lack of suitable immunohistochemical markers to distinguish between follicular thyroid carcinoma and other types of follicular-derived lesions complicates diagnosis, and despite most of these tumours being surgically resected, only a small number will test positive for malignancy. As such, the development of new orthogonal diagnostic approaches may improve the accuracy of diagnosing thyroid nodules.

DESIGN

This study includes a retrospective, multi-centre training cohort including 54 fresh-frozen follicular-patterned thyroid samples and two independent, multi-centre validation cohorts of 103 snap-frozen biopsies and 33 FNAC samples, respectively.

METHODS

We performed a genome-wide genetic and epigenetic profiling of 54 fresh-frozen follicular-patterned thyroid samples using exome sequencing and the Illumina Human DNA Methylation EPIC platform. An extensive validation was performed using the bisulfite pyrosequencing technique.

RESULTS

Using a random forest approach, we developed a three-CpG marker-based diagnostic model that was subsequently validated using bisulfite pyrosequencing experiments. According to the validation cohort, this cost-effective method discriminates between benign and malignant nodules with a sensitivity and specificity of 97 and 88%, respectively (positive predictive value (PPV): 0.85, negative predictive value (NPV): 0.98).

CONCLUSIONS

Our classification system based on a minimal set of epigenetic biomarkers can complement the potential of the diagnostic techniques currently available and would prioritize a considerable number of surgical interventions that are often performed due to uncertain cytology.

SIGNIFICANCE STATEMENT

In recent years, there has been a significant increase in the number of people diagnosed with thyroid nodules. The current challenge is their etiological diagnosis to discount malignancy without resorting to thyroidectomy. The method proposed here, based on DNA pyrosequencing assays, has high sensitivity (0.97) and specificity (0.88) for the identification of malignant thyroid nodules. This simple and cost-effective approach can complement expert pathologist evaluation to prioritize the classification of difficult-to-diagnose follicular-patterned thyroid lesions and track tumor evolution, including real-time monitoring of treatment efficacy, thereby stimulating adherence to health promotion programs.

A prognostic microRNA-based signature for localized clear cell renal cell carcinoma: the Bio-miR study

BACKGROUND

Two thirds of renal cell carcinoma (RCC) patients have localized disease at diagnosis. A significant proportion of these patients will relapse. There is a need for prognostic biomarkers to improve risk-stratification and specific treatments for patients that relapse. The objective of this study is to determine the clinical utility of microRNA signatures as prognostic biomarkers in localized clear cell RCC (ccRCC) and propose new therapeutic targets in patients with a high-risk of relapse.

PATIENTS AND METHODS

The microRNA profiles from a discovery cohort of 71 T1-T2 ccRCC patients (n = 88) were analyzed using microarrays. MicroRNAs prognostic value was established, and a microRNAs signature predicting relapse for T1b-T3 disease was defined. Independent validation was carried out by qPCR in cohorts from UK (n = 75) and Spain (n = 180), and the TCGA cohort (n = 175). In the Spanish validation cohort, proteomics experiments were done. Proteins were extracted from FFPE tissue and analyzed using by data-independent acquisition mass spectrometry. Additionally, ccRCC TCGA RNA-seq data was also analyzed. Both protein and RNA-seq data was analyzed using Significance Analysis of Micorarrays (SAM) and probabilistic graphical models, which allow the identification of relevant biological processes between low and high-risk tumors.

RESULTS

A 9-microRNAs signature, Bio-miR, classified patients into low- and high-risk with disease-free survival (DFS) at 5 years of 87.12 vs. 54.17% respectively (p = 0.0086, HR = 3.58, 95%CI: 1.37-8.3). Results were confirmed in the validation cohorts with 5-year DFS rates of 94% vs. 62% in the UK cohort (HR = 7.14, p = 0.001), 82.9% vs. 58.7% in the Spanish cohort (HR = 2.46, p = 0.0013), and 5-year overall survival rates of 72.7% vs. 44.5% in the TCGA cohort (HR = 2.43, p = 0.0012). Among low-risk patients according to adjuvant immunotherapy clinical trial criteria, Bio-miR identified a high-risk group. Maybe those patients ought to be considered to receive adjuvant therapy. Proteins overexpressed in the high-risk group were mainly related to focal adhesion, serine and inositol metabolism, and angiogenesis. Probabilistic graphical models defined eight functional nodes related to specific biological processes. Differences between low- and high-risk tumors were detected in complement activation and translation functional nodes. In ccRCC TCGA cohort, 676 genes were differentially expressed between low and high-risk patients, mainly related to complement activation, adhesion, and chemokine and cytokine cascades. In this case, probabilistic graphical models defined ten functional nodes. Calcium binding, membrane, adhesion, extracellular matrix, blood microparticle, inflammatory response and immune response had higher functional node activity, and metabolism node, containing genes related to retinol and xenobiotic and CYP450 metabolism, had lower activity in the high-risk group.

CONCLUSIONS

Bio-miR dichotomizes ccRCC patients with non-metastatic disease into those with low- and high-risk of relapse. This has implications for treatment and follow-up, identifying patients most likely to benefit from adjuvant treatment in clinical trials, preventing unnecessary exposure to side-effects, and providing health economics benefits. Additionally, promising therapeutic targets, as angiogenesis, immune response, metabolism, or complement activation, were found deregulated in high-risk ccRCC patients defined by Bio-miR. These findings may be useful to select patients for tailored, molecularly-driven clinical trials. Identifying which patients with kidney cancer are most at risk of their cancer coming back after surgery is critical, so that they can be prioritized for early treatment. We have identified a combination of biomarkers present in the cancer tissue (called BiomiR) which can help to do this.