Characterization of Subtypes of BRAF-Mutant Papillary Thyroid Cancer Defined by Their Thyroid Differentiation Score

Abstract

Context

The BRAF^V600E mutation has been associated with more advanced clinical stage in papillary thyroid cancer (PTC) and decreased responsiveness to radioiodine (RAI). However, some BRAF mutant PTCs respond to RAI and have an indolent clinical behavior suggesting the presence of different subtypes of BRAF mutant tumors with distinct prognosis.

Objective

To characterize the molecular and clinical features of 2 subtypes of BRAF-mutant PTCs defined by their degree of expression of iodine metabolism genes.

Design

227 BRAF-mutant PTCs from the Cancer Genome Atlas Thyroid Cancer study were divided into 2 subgroups based on their thyroid differentiation score (TDS): BRAF-TDS^hi and BRAF-TDS^lo. Demographic, clinico-pathological, and molecular characteristics of the 2 subgroups were compared.

Results

Compared to BRAF-TDS^hi tumors (17%), BRAF-TDS^lo tumors (83%) were more frequent in blacks and Hispanics (6% vs 0%, P = 0.035 and 12% vs 0%, P = 0.05, respectively), they were larger (2.95 ± 1.7 vs 2.03 ± 1.5, P = 0.002), with more tumor-involved lymph nodes (3.9 ± 5.8 vs 2.0 ± 4.2, P = 0.042), and a higher frequency of distant metastases (3% vs 0%, P = 0.043). Gene set enrichment analysis showed positive enrichment for RAS signatures in the BRAF-TDS^hi cohort, with corresponding reciprocal changes in the BRAF-TDS^lo group. Several microRNAs (miRs) targeting nodes in the transforming growth factor β (TGFβ)-SMAD pathway, miR-204, miR-205, and miR-144, were overexpressed in the BRAF-TDS^hi group. In the subset with follow-up data, BRAF-TDS^hi tumors had higher complete responses to therapy (94% vs 57%, P < 0.01) than BRAF-TDS^lo tumors.

Conclusion

Enrichment for RAS signatures, key genes involved in cell polarity and specific miRs targeting the TGFβ-SMAD pathway define 2 subtypes of BRAF-mutant PTCs with distinct clinical characteristics and prognosis.

The BRAF^V600E mutation (henceforth referred to as BRAF), which constitutes ~60% of all driver mutations in papillary thyroid cancer (PTC), has received particular attention since it has been associated with a higher prevalence of nodal metastases and a more advanced clinical stage (1,2). The combination of BRAF mutations with TERT promoter mutations has been linked with more aggressive behavior of tumors and worse clinical outcomes including slightly higher mortality (3,4); however, the prognostic and predictive power of BRAF^V600E alone is less certain since it has not proven to be independently associated with worse outcomes (2,5-7). Other genetic and clinico-pathological unfavorable characteristics in combination with BRAF appear to confer a worse prognosis (8,9). Despite the fact that BRAF has been shown to downregulate genes critical for radioactive iodine responsiveness, such as the thyroid-stimulating hormone receptor, sodium-iodide symporter, and thyroglobulin (10-12), some BRAF mutant tumors do appear to respond to radioactive iodine (13,14). This suggests that subsets of BRAF mutant PTCs may have diverse differentiation states and clinical behaviors.

The purpose of this study was to better characterize 2 subtypes of BRAF-mutant PTCs based on their expression of iodine metabolism genes in the Cancer Genome Atlas (TCGA) study (15), characterize their molecular and clinical features and ascertain differential prognosis. We found that the thyroid differentiation score (TDS) defined BRAF-mutant PTC groups with distinct TNM status and clinical outcomes. Intriguingly BRAF-mutant PTCs with relatively preserved expression of thyroid differentiation genes (BRAF-TDS^hi PTCs) had higher expression of a set of microRNAs (miRs) that target messenger RNAs (RNAs) encoding for nodes in the transforming growth factor β (TGFβ) signaling pathway, which when activated dampens thyroid-specific gene expression.

Methods

Population

We evaluated the clinical and pathologic characteristics of the 227 BRAF-mutant PTCs from the TCGA Thyroid Cancer (THCA) study (15). We defined 2 subgroups of BRAF^V600E mutant tumors based on their TDS, an integrated measure of expression of 16 iodine metabolism genes defined by TCGA (15): (1) PTCs with relatively preserved expression of thyroid differentiation genes (BRAF-TDS^hi; TDS > 0) and (2) PTCs with decreased expression of thyroid differentiation genes (BRAF-TDS^lo; TDS: < 0). The clinical and pathologic characteristics of the tumors were extracted from the TCGA database (15). TCGA included all the appropriate variables to classify patients into low-, intermediate-, and high-risk groups according to the American Thyroid Association (ATA) risk stratification system (16).

As there is limited follow-up information on the TCGA THCA patient population, we examined a subset of 54 patients with BRAF-mutant PTCs that Memorial Sloan Kettering Cancer Center (MSK) had contributed to TCGA and on whom we had follow-up clinical data. To determine whether patients with BRAF-TDS^hi and BRAF-TDS^lo had different clinical outcomes, we mirrored the ATA guidelines response to therapy stratification system (16) and defined as “complete responders” those patients without structurally identifiable disease and with negative thyroglobulin and thyroglobulin antibodies. “Incomplete responders” were those with either structural disease on cross-sectional imaging, biochemical persistence of thyroglobulin, or increasing thyroglobulin antibodies over time or nonspecific findings on imaging studies whether they were later proven to be metastatic thyroid cancer. Since most tumors collected in the TCGA project were ATA low risk, only 18 out of 54 patients had radioiodine therapy after surgery. We validated our findings with 46 BRAF-mutant PTCs that MD Anderson Cancer Center (MDA) had contributed to TCGA and on whom we had follow-up clinical data. Our study was reviewed and approved by the Institutional Review Board at MSK (IRB# 15-082) and by the Institutional Review Board at MDA (IRB# LAB06-1007).

Differential Gene Expression

The list of iodine metabolism genes that TCGA prioritized to create the TDS were DIO1, DIO2, DUOX1, DUOX2, FOXE1, GLIS3, PAX8, NKX2-1, SLC26A4, SLC5A5, SLC5A8, TG, TPO, TSHR, THRA, and THRB. The TDS for each sample is the average of the log-fold change in expression of every gene in the set as compared to its respective median derived from the entire cohort. The BRAF-RAS score (BRS) was derived as described in the THCA TCGA study (15). We investigated whether concomitant somatic mutations as measured in THCA TCGA were differentially represented in the 2 BRAF-mutant groups. To identify additional genes differentially expressed between the 2 BRAF subgroups we compared the expression of 20 531 genes and miRs sequenced with RNASeq in TCGA between the BRAF-TDS^hi and BRAF-TDS^lo groups. We also used the TDS score to determine whether it predicted response to therapy among BRAF-mutant patients.

Statistical Methods

We extracted TDS and BRS scores for every sample from the TCGA database. Clinical variables were compared with Student t test and Mann-Whitney test for continuous variables and with Chi-squared test for categorical variables. A 2-tailed P-value less than 0.05 was considered statistically significant for all calculations. We conducted differential gene expression analysis as well as gene set enrichment analysis (GSEA) between the BRAF TDS^hi and TDS^lo subgroups. For both analyses, we used the normalized counts from RNAseq as the measure of gene expression. We used an expression filter to remove genes with low expression or low variability by using a threshold for the minimum normalized counts as well as median absolute deviation threshold. Gene expression between the 2 groups was compared using Wilcoxon-Mann-Whitney rank sum tests. We used STATA 12 (College Station, TX, USA) and R (R Core Team 2021, Vienna, Austria, https://www.R-project.org/).

For the differential expression analysis, we used all 400 samples that underwent RNAseq expression profiling. We used 2 filtering criteria: a stringent one, selecting for transcripts with a normalized read count ≥ 10 in at least 300 samples and a median absolute deviation (in log2 scale) of 0.5, which resulted in 6306 of the 20 531 transcripts being considered for analysis, and a more permissive filtering with normalized read count ≥ 5 in at least 150 samples and a median absolute deviation of at least 0.3, which resulted in 12 925 transcripts being considered. The normalized read counts were compared between the 2 groups using Wilcoxon-Mann-Whitney rank sum test. We used volcano plots to show the P-values plotted against the fold-change for the transcripts under consideration. We used the GSEA tool to further explore pathways that could have contributed to the differential expression of genes in the 2 groups.

Results

A total of 391 (of 496) PTCs from the THCA TCGA had both whole exome sequencing and RNAseq data (15). Of these, 227 tumors harbored a BRAF^V600E mutation (58%). We divided the BRAF-mutant tumors into 2 groups based on the integrated expression levels of thyroid differentiation genes as reflected in their TDS: BRAF-TDS^hi and BRAF-TDS^lo. The BRAF-TDS^hi group constituted 17.2% of the total (39/227) (Fig. 1). Although by definition this group had a positive TDS (0.64 ± 0.48), it was significantly lower than that of RAS-mutant PTCs (1.2 ± 0.53), which tend to be more responsive to radioactive iodine (13,14).

Figure 1.

Association of genotype with expression of genes regulating thyroid differentiated function in the Cancer Genome Atlas study of papillary thyroid carcinoma. Each column represents a tumor from an individual patient. Y axis: Top: Oncoprint of the most prevalent driver mutations, with BRAF-V600E representing 61% of the sample. Middle: RNAseq of thyroid differentiation genes, representing messenger RNAs (mRNAs) involved in thyroid hormone biosynthesis and iodine metabolism. Bottom: Differentially expressed miRs between genotypes. Bar at foot of figure shows color-coded log2 (fold change) mRNA and microRNA expression levels. Based on Figure 5 of (15).

We explored the clinical and pathologic characteristics of the 2 subgroups. Table 1 shows that there were no differences in age or sex between them. However, there was a significant difference in race and ethnicity, with blacks and Hispanics being overrepresented in the BRAF-TDS^lo group (6% vs 0%, P = 0.035 and 12% vs 0%, P = 0.05, respectively). Asians made up to a fifth (21%) of the BRAF-TDS^hi group. There were no differences between the groups in terms of underlying thyroid pathology (P = 0.34), history of radiation (P = 0.60), or family history of thyroid cancer (P = 0.39), although for the latter data were not available in 95% of patients (Table 1).

Table 1.

Clinical characteristics of BRAF mutant tumors by expression of iodine metabolism genes

	Clinical Characteristics (n = 227)
	BRAF-TDSlo (n = 188)	BRAF-TDShi (n = 39)	P-value
Age at diagnosis, years, mean ± SD	47.5 ± 15.5	49.1 ± 15.1	0.55
Race			0.035
White	139 (74)	23 (59)
Black	11 (6)	0 (0)
Asian	12 (6)	8 (21)
Not available	20 (10)	6 (15)
Unknown	6 (3)	2 (5)
Ethnicity			0.05
Hispanic	22 (12)	0 (0)
Not Hispanic	139 (74)	32 (82)
Not available	18 (10)	4 (10)
Unknown	9 (5)	3 (8)
Sex			0.78
Female	139 (74)	28 (72)
Male	49 (26)	11 (28)
Past medical history			0.34
Hypothyroidism	9 (5)	1 (3)
LT	19 (10)	5 (13)
NH	20 (11)	8 (21)
LT + NH	9 (5)	3 (8)
Normal	112 (60)	17 (44)
Not available	17 (9)	3 (8)
Prior thyroid cancer	14 (7.5)	2 (5)
Radiation exposure			0.60
No	159 (85)	35 (90)
Yes	8 (4)	0 (0)
First degree relative with TC			0.39
Parent	3 (1.6)	2 (5)
Child	1 (0.5)	0 (0)
Sibling	4 (2.13)	0 (0)
Not available	180 (96)	37 (95)

Data are given as n (%) unless otherwise noted.

Abbreviations: BRAF-TDS^hi, thyroid differentiation score > 0; BRAF-TDS^lo, thyroid differentiation score < 0; LT, lymphocytic thyroiditis; NH, nodular hyperplasia; TC, thyroid cancer.

Tumor characteristics differed significantly between the 2 groups. Interestingly, tumors in the group with downregulation of iodine metabolism genes (BRAF-TDS^lo) were larger (P = 0.002), had a higher pathologic T stage (P = 0.002), more tumor-involved lymph nodes (P = 0.042), and a higher frequency of distant metastatic disease (P = 0.043) (Table 2). We assessed vital status based on the TCGA data and did not find a significant survival difference between the groups (97% in BRAF-TDS^hi vs 96% in BRAF-TDS^lo alive, P = 0.62).

Table 2.

Tumor characteristics of BRAF-mutant samples by expression of iodine metabolism genes

	Tumor characteristics (n = 227)
	BRAF-TDSlo (n = 188)	BRAF-TDShi (n = 39)	P-value
Histological type			0.40
PTC, classical variant	154 (82)	31 (80)
PTC, follicular variant	9 (5)	4 (10)
PTC, tall cell variant	23 (12)	4 (10)
Side			0.32
Right	70 (37)	13(33)
Left	68 (36)	17 (44)
Isthmus	13 (7)	0 (0)
Bilateral	33 (18)	9 (23)
Focality			0.73
Unifocal	100 (53)	21 (54)
Multifocal	85 (45)	18 (46)
Size, max diam in cm, mean ± SD	2.95 ± 1.7	2.03 ± 1.5	0.002
Residual tumor			0.18
R0	139 (74)	35 (90)
R1	22 (12)	2 (5)
R2	3 (1.6)	0 (0)
Rx	14 (7)	1 (3)
Extrathyroidal extension			0.14
Minimal T3	68 (36)	11 (28)
Moderate T4	11 (6)	0 (0)
None	104 (55)	28 (72)
Pathologic T			0.002
T1	12 (6)	5 (13)
T1a	4 (2)	5 (13)
T1b	28 (15)	9 (23)
T2	56 (30)	7 (18)
T3	75 (40)	12 (31)
T4	5 (3)	0 (0)
T4a	8 (4)	0 (0)
Pathologic N			0.002
N0	64 (34)	26 (67)
N1	27 (14)	0 (0)
N1a	48 (26)	6 (15)
N1b	31 (17)	5 (13)
Nx	18 (10)	2 (5)
Number of positive lymph nodes ± SD	3.9 ± 5.8	2.0 ± 4.2	0.042
Pathologic M			0.043
M0	105 (56)	30 (77)
M1	5 (3)	0 (0)
Mx	78 (42)	9 (23)
Pathologic stage			0.16
I	95 (51)	27 (69)
II	16 (8.5)	1 (3)
III	47 (25)	10 (26)
IVa	25 (13)	1 (3)
IVc	4 (2)	0(0)
ATA risk			0.20
Low	88 (47)	23 (59)
Intermediate	91 (48)	16 (41)
High	9 (5)	0
Vital status			0.62
Alive	180 (96)	38 (97)
Dead	8 (4)	1 (3)

Data are given as n (%) unless otherwise noted.

Abbreviations: ATA, American Thyroid Association; BRAF-TDS^hi: thyroid differentiation score > 0; BRAF-TDS^lo: thyroid differentiation score < 0.

TERT promoter mutations were shown to anticorrelate with TDS in the TCGA cohort as a whole. Despite the low prevalence of TERT promoter mutations in papillary thyroid carcinoma (9.4%), 24 of 183 BRAF-TDS^lo patients had a TERT promoter mutation whereas only 2 of 37 BRAF-TDS^hi patients were TERT mutant (13% vs 5.4%, P = 0.146).

GSEA of the TCGA transcriptomes using the Molecular Signature Database (MSigDB) database showed positive enrichment for RAS signatures in the TCGA BRAF-TDS^hi cohort (P = 0.022; normalized enrichment score = 1.977), with corresponding reciprocal changes in the BRAF-TDS^lo group (P = 0.0074; normalized enrichment score = −1.83). This is consistent with their BRSs, which showed that the TDS^hi tumors were more RAS-like. Furthermore, the extracellularly regulated kinase output score was significantly lower in the BRAF-TDS^hi vs TDS^lo tumors (median 11.3 vs 15.5, P = 0.018), suggesting a tempered mitogen-activated protein kinase (MAPK) output in the better differentiated group. Volcano plots showed 475 genes differentially expressed between the groups, primarily through overexpression in BRAF-TDS^hi vs TDS^lo (Fig. 2). As expected, several TDS genes (TPO, DIO1) and TDS-associated genes (MTIG) were among those expressed with a > 3 log-fold change. Other genes of interest overexpressed at these levels included CDH16, which encodes the adherens junction protein cadherin 16. CDH16 is required for thyroid follicular cell polarity (17), a critical step in the assembly of the thyroid follicle, the functional unit of the thyroid gland. Based on this, we examined the proportion of follicular variant PTCs in the 2 groups and interrogated the TCGA data set for possible differences in the follicular fraction (FF), a histopathological measure of the fraction of the tumor specimen consisting of cancer cells with a preserved follicular architecture (15). The median FF was 10 for TDS^lo and 15 for TDS^hi groups (Wilcoxon P-value = 0.052), showing a trend toward retention of follicular structures in the TDS^hi cohort. Follicular variant PTC is mostly associated with RAS mutations, but a few BRAF mutant tumors in TCGA were found to harbor this variant. Within these, follicular variant PTCs occurred more frequently in the subgroup with preserved expression of iodine metabolism genes (BRAF-TDS^hi), but this difference was not statistically significant given the low absolute numbers in each group (n = 4/39 BRAF-TDS^hi, 10% vs n = 9/188 BRAF-TDS^lo, 5%, P = 0.4).

Figure 2.

Differential gene expression between BRAF-TDS^hi and BRAF-TDS^lo papillary thyroid cancers (PTCs): volcano plot of differentially expressed genes between BRAF-TDS^hi and BRAF-TDS^lo tumors from the Cancer Genome Atlas study of PTC. Labeled genes with a log-fold change > 3 are shown. Green and blue dots represent messenger RNAs with a P < 0.05 using the Benjamini-Hochberg false discovery rate correction. Blue dots represent transcripts differentially expressed at a multiple testing adjusted P-value < 0.05 using the more stringent Bonferroni method to determine the family wide error rate.

PDHD1-L1 encodes for fibrocystin-L1, a protein expressed in primary cilia, mutations of which confer predisposition to polycystic kidney disease. Its expression was increased in BRAF-TDS^hi tumors by ~6 log-fold compared to BRAF-TDS^lo. Primary cilia may also play a role in planar cell polarity, and PDHD1-L1 has been proposed to function as a tumor suppressor in thyroid cancer based on its decreased expression in PTCs and on induction of cell growth following gene silencing experiments in thyroid cancer cell lines (18).

TFF3 encodes for trefoil factor 3, and it is expressed at a >5 log-fold higher levels in the TDS^hi cohort. Trefoils are a family of secreted proteins from mucin-producing cells. TFF3 has been studied as a diagnostic biomarker to distinguish benign from malignant thyroid neoplasms, since its expression is lower in papillary and follicular carcinomas compared to follicular adenomas (19).

A volcano plot showing the differential expression of miRs is shown in Figure 3. Among these, miR-136, miR-138, miR-204, miR-205, and miR-144 were the highest differentially expressed genes. Interestingly, several of the overexpressed miRs target nodes in the TGFβ signaling pathway. miR 204-5p was among the highest differentially expressed genes, and it has been reproducibly reported to target TGFβR1 and TGFβR2 mRNAs as well as SNAIL, an EMT marker induced by this pathway (20-22). miR 205 and 144, also overexpressed in BRAF TDS^hi tumors, target TGFβR2 and SMAD4, respectively (23,24). As the TGFβ-SMAD pathway decreases thyroid differentiation (including most genes in TDS), overexpression of these miRs could contribute to their higher TDS scores.

Figure 3.

Differential expression of microRNA genes between BRAF-TDS^hi and BRAF-TDS^lo papillary thyroid cancers (PTCs): volcano plot of differentially expressed microRNAs (miRs) between BRAF-TDS^hi and BRAF-TDS^lo PTCs. miRs with a log-fold change > 1 are labeled.

miR-136-5p has been shown to decrease the expression of metadherin, a transmembrane protein that has been associated with advanced TNM stage and lymph node and distant metastases in thyroid carcinoma (25). Metadherin influences several oncogenic signaling pathways and transcription factors, such as the Ras, Myc, phosphatidylinositol 3-kinase/AKT, nuclear factor-κB, MAPK, and Wnt pathways (26,27). miR-138 overexpression in thyroid cancer has been shown to inhibit long noncoding RNA RP11-476D10.1, resulting in decrease cell proliferation and apoptosis in vitro and decreased tumor formation in vivo (28). Moreover, silencing of long noncoding RNA H19 inhibited invasion and migration of PTC cells via miR-138-dependent regulation of Leucine-rich repeat kinase 2 (29).

To evaluate response to therapy, we examined a cohort of 54 BRAF-mutant patients that MSK contributed to TCGA and on which we had adequate clinical follow-up. Eighteen of 54 patients belonged to the BRAF-TDS^hi group (33%) and 36 of 54 patients had underexpression of iodine metabolism genes (BRAF-TDS^lo; 67%). We found that 15/16 (94%) BRAF-TDS^hi patients were complete responders to therapy, and 1 had a biochemical incomplete response. In the BRAF-TDS^lo group, 20/35 patients (57%) had a complete response to therapy, and 43% were incomplete responders. Both the TDS as well as the BRS were significantly higher among complete responders than among incomplete responders (−0.3 vs −0.9, P = 0.023 and −0.8 vs −0.96, P = 0.032) (Fig. 4A). Only 18/54 patients were treated with RAI (13 BRAF-TDS^lo and 5 BRAF-TDS^hi), so this cohort was not powered to look at the impact of TDS on RAI response. Only 2 patients had uptake outside of the thyroid bed that persisted as structural disease over time and both belonged to the BRAF-TDS^lo subgroup. One of 5 BRAF-TDS^hi patients (20%) treated with RAI had a biochemical incomplete response to therapy compared to 7 out of 13 (54%) in the BRAF-TDS^lo group who had biochemically and/or structurally incomplete responses to therapy over a median follow-up of 6.2 years (P = 0.19).

Figure 4.

Association of thyroid differentiation score (TDS) and BRAF-RAS score with response to therapy: (A) TDS and BRAF-RAS scores of BRAF-mutant papillary thyroid cancers from Memorial Sloan Kettering Cancer Center (n = 54) (A) and from MD Anderson Cancer Center (n = 46) (B) based on response to therapy. See text for definition of complete and incomplete responders.

To validate these findings, we collected clinical follow-up information on 46 patients with BRAF-mutant tumors that MDA had contributed to TCGA, calculated their TDS, and correlated it with response to therapy. Overall, 5/46 (11%) patients were BRAF-TDS^hi and 1/5 patients had persistent disease at final follow-up. Thirteen out of 41 (32%) BRAF-TDS^lo tumors were incomplete responders at final follow-up. Overall, TDS was higher among complete responders than among incomplete responders, but this was not statistically significant (−0.84 vs −1.14, P = 0.11). The BRS was not associated with response to therapy in the MDA BRAF cohort (Fig. 4B). In contrast to the MSK cohort, 35/46 (76%) patients of the MDA cohort received RAI therapy after surgery. None of the 4 patients with BRAF-TDS^hi who received RAI developed persistent disease, whereas 8 of 29 (28%) BRAF-TDS^lo patients who had received RAI were incomplete responders at final follow-up (P < 0.01). The area under the curve for TDS in terms of prediction of response to therapy was 0.639 for MSK and 0.662 for MDA.

Combined analysis of the MSK and MDA cohorts show that 2/21 (9.5%) BRAF-TDS^hi patients had an incomplete response to therapy compared to 28/76 (37%) in the BRAF-TDS^lo cohort (P < 0.01).

Discussion

In this study we show that BRAF mutant tumors can be subclassified into 2 distinct subgroups: one with preserved expression of iodine metabolism genes (BRAF-TDS^hi, ~20%) where tumors are smaller, with less tumor-involved lymph nodes, and a second subgroup with downregulation of iodine metabolism genes (BRAF-TDS^lo, ~80%), more commonly seen in blacks and Hispanics, with larger tumors, more tumor-involved lymph nodes, and a higher frequency of distant metastatic disease. The explanation for these clinical differences has been unexplored so far.

Our study found 3 lines of molecular evidence that may help distinguish these subtypes of BRAF mutant tumors. First, the transcriptomes of the BRAF-TDS^hi tumors showed them to be more RAS-like based on their BRSs. The extracellularly regulated kinase output scores were lower in BRAF-TDS^hi tumors compared to BRAF-TDS^lo tumors, indicating that the MAP kinase pathway appears to be tempered in the more differentiated BRAF tumors.

Second, we found differentially expressed genes that are involved in cell polarity and are essential for the preservation of the structure of the thyroid follicle. CDH16, which encodes an adherens protein required for cell polarity that is essential for thyroid cell function and thyroid hormone biosynthesis, was overexpressed in the BRAF-TDS^hi group. In addition, the overall FF of BRAF-TDS^hi tumors was higher than that of BRAF-TDS^lo PTCs. Taken together, these findings suggest that a preserved follicular architecture with partially maintained differentiated function are key features of the BRAF-TDS^hi group.

Third, the TCGA THCA study showed that the expression of certain miRs, including miR-146b-3p, miR-146b-5b, miR-375, miR-204-5p, miR-21-5p, and miR-7-5p, clustered with more differentiated tumors. Interestingly, the miRs overexpressed in BRAF-TDS^hi PTCs compared to the BRAF-TDS^lo group were different from those previously reported to correlate with TDS across all PTC genotypes (15). miR-204 has been reported to target TGFβR1 and TGFβR2 mRNAs (20,22) as well as SNAIL (21), a marker of epithelial-to-mesenchymal transition induced by this pathway. As the TGFβ-SMAD pathway decreases thyroid differentiation and downregulates most genes involved in TDS, overexpression of miR-204 in BRAF-TDS^hi tumors could contribute to their higher TDS. Similarly, miR-205 and miR-144 were also overexpressed in BRAF-TDS^hi tumors and target SMAD4 and TGFβ2, respectively. Overexpression of miR-205 directly binds to the 3’-UTR of SMAD4 suppressing its expression, dampening TGFβ signaling output (23). miR-144 suppresses the invasion and migration capability of thyroid cancer and suppress the expression of zinc finger E-box-binding homeobox 1 and 2, the 2 TGFβ-regulated E-cadherin suppressors, by directly binding their 3’UTRs (30). The potential relevance of these findings is highlighted by a recent study showing that blockade of TGFβ and activin-induced SMAD signaling cooperated with MAPK inhibitors to restore iodine incorporation in murine Braf^V600E-PTCs (31).

TDS was associated with response to therapy in 54 BRAF mutant patients from MSK, and there was a trend toward this finding among a validation cohort of 46 patients from MDA. No unsupervised gene or set of genes was significantly associated with response to therapy when adjusted for multiple testing. The power of TDS to predict response to therapy was relatively low (area under the curve: 0.639 and 0.662) given that most tumors collected in TCGA were early stage and would have probably been cured with surgery alone.

We have shown that despite their genomic simplicity BRAF mutant tumors are heterogeneous and that this heterogeneity is not dictated by concomitant somatic mutations. Instead, the TDS defines distinct subgroups that are clinically, pathologically, and molecularly different. We could not establish whether the TDS categorization identified patients that may respond differentially to RAI given that in general the tumors were low risk and most patients were treated with surgery alone. This remains a clinically relevant and potentially tractable question given that several genes and miRs differentially expressed between the groups are involved in iodine metabolism pathways.

Additional Information

Disclosure: The authors have nothing to disclose.

Data Availability

Some data generated or analyzed during this study are included in the data repositories listed in the references and some data related to response to therapy calculations that were generated and analyzed during the current study are not publicly available but are available from the corresponding author on reasonable request.