Abstract Context Although 60% of papillary thyroid carcinomas are BRAFV600E mutant (PTCV600E), the increased aggressiveness of these cancers is still debated. Objective For PTCV600E we aimed to further characterize the extent of the stroma and its activation, the three-dimensional (3D) tumor-stroma interface, and the proliferation rates of tumor and stromal fibroblasts. Design We analyzed exomes, transcriptomes, and images of 364 papillary thyroid carcinoma (PTCs) from The Cancer Genome Atlas (TCGA), including 211 PTCV600E; stained 22 independent PTCs for BRAFV600E and Ki67; sequenced the exomes and stained BRAFV600E in 5 primary tumor blocks and 4 nodal metastases from one patient with PTCV600E; and reconstructed the 3D volumes of one tumor and one metastatic block at histological resolution. Results In TCGA, BRAFV600E was associated with higher expression of proliferation markers and lower expression of thyroid differentiation markers, independently of tumor purity. Moreover, PTCV600E, in line with their overall lower purity, also had higher expression of fibroblast- and T cell–associated genes and presented more fibrosis. Tumor cells that appeared disconnected on two-dimensional histological slices were revealed to be part of a unique tumor component in the 3D reconstructed microvolumes, and they formed a surprisingly complex connected space, infiltrating a proliferative stroma. Finally, in our PTC set, both stromal fibroblasts and tumor cells presented higher proliferation rates in PTCV600E. Conclusions Our results support the increased aggressiveness associated with BRAFV600E in PTC and shed light on the important role of the stroma in tumor expansion. The greater and more active fibrotic component predicts better efficiency of combined targeted treatments, as previously proposed for melanomaV600E. To focus on informative sequencing data, most oncogenomic studies reject samples with low tumor cell content based on pathology examination. Nevertheless, low-purity samples routinely end up being sequenced and in most cases are deemed useless for further analysis. This paper stems from the sequencing of such series of tumor regions, originally collected for a study of the intratumor and interfoci heterogeneity in papillary thyroid carcinoma (PTC). We reasoned that further investigation of the underlying tumor morphology could shed light on the contribution to tumor expansion of noncancerous cells. The shape of the tumor mass is important because it defines the contact area with the stroma, where features of aggressiveness, such as invasiveness and increased proliferation, are observed (1). Some of these carry prognostic information in various cancer types, such as oral cancers (2). In fact, together with the morphology of the tumor cells, the morphology of the tumor mass relates to the mode of metastatic dissemination (3). Cancer-associated fibroblasts (CAFs) pull on tumor cells, generating tumor elongations (4). Interestingly, many PTCs present poorly defined boundaries (5). Liu et al. (6) showed an association between PTC recurrence and presence of isolated clusters of cells and loss of cell polarity [i.e., two features of epithelial-to-mesenchymal transition (EMT)]. Moreover, microPTCV600E (PTC <1 cm, with BRAFV600E mutation) seemed to present invasive fronts with a more complex architecture, encompassing markedly invasive contours, when compared with microPTCWT (7). Isolated clusters of tumor cells at the invasive front could be interpreted as the consequence of EMT, followed by a return to the epithelial state (1). Alternatively, these isolated islands of cells observed on two-dimensional (2D) slices could also belong to the same connected three-dimensional (3D) tumor volume, implying branching patterns drifting from an ellipsoid. So far, 3D tumor volumes have been mostly acquired by medical imaging techniques, such as computed tomography and magnetic resonance imaging, common in the field of radiomics (8). These technologies, which operate at macroscopic scale, typically yield ellipsoid-shaped tumor volumes with fingering patterns. In this context the tumor-stroma interface is implicitly assumed to sit on an ellipsoid periphery. For example, Vasko and collaborators (9) analyzed differential expression between peripheral and central parts of PTC, suggesting a higher degree of EMT at the periphery. The assumption that the tumor-stroma contact surface is ellipsoid, however, has never been demonstrated at the fine scale relevant to cell-to-cell contact. Upon zooming in on the invasive front with higher-resolution imaging, the observed tumor-stroma contact area may increase quite a lot, akin to coastline lengths at different magnifications in the coastline paradox (10). The fractal dimension measures the increase in tumor-stroma contact area in proportion to the magnification. At histological resolution, the morphology of the tumor mass has been described as complex and the fractal dimension was linked to prognosis (11). In colorectal carcinoma, tumor budding is a complex pattern whereby isolated clusters of cells at the invasive front have undergone EMT, and it has been associated with worse prognosis (12). In our study, we explored the 3D shape of the tumor/stroma contact area at a histological scale. PTCs represent ∼80% of all thyroid cancers (13) and, given the rapidly increasing incidence of microPTC, mostly due to earlier diagnosis, thyroid cancer could become the fourth most diagnosed cancer in the United States by 2030 (14). A recent study by The Cancer Genome Atlas (TCGA) (15) identified driver mutations in 96.5% of 496 PTCs. In line with previous studies, PTC progression was essentially driven by constitutive activation of the mitogen-activated protein kinase (MAPK) pathway, with the most frequent mutations found in the BRAF, RAS, and RET genes (15, 16). BRAFV600E was the most frequent mutation, detected in ∼60% of PTCs (15, 17). Although several studies have found an association of BRAFV600E with clinical features of aggressiveness and bad prognosis in PTC (7, 18), this has often been disputed (19). Most notably, BRAFV600E was not associated with distant metastasis (20, 21). However, in PTCV600E presenting poorly differentiated or undifferentiated areas, it was consistently detected in both regions, suggesting a role in the progression from PTC to the aggressive undifferentiated cancers (17, 22). Here, the analysis of TCGA histological images and genomic and transcriptomic data shows that our low-purity multiregion PTC is part of a continuum of PTC structures in which BRAFV600E PTC stands out with strong CAF activation, lower tumor cell purity, dedifferentiation of tumor cells, and a high proliferation rate in both the tumor cell and the CAF compartments. Materials and Methods PTCV600E case description A 51-year-old woman diagnosed with a bilateral PTC (right lobe: 0.6 cm; left lobe: 2.8 cm) and concurrent metastatic involvement of the lymph nodes (stage IVA, pT3N1b) was treated with total thyroidectomy and lymphadenectomy (five nodal metastases in the recurrent and mediastinal areas). Iodine-131 radiotherapy was administered 1 month after surgery. Twelve months after thyroidectomy, nodal recurrences were removed (four nodal metastases from the left cervical, three of which were included in the study: M2, M3, and M4). First histopathological examination revealed a classic PTC in a goiter context and suggested a tumor cell fraction >70%. Tissue samples For immunohistochemistry (IHC) and exome analyses, tumor and normal thyroid tissues were obtained from the Jules Bordet Institute tissue bank (Brussels, Belgium). We included four foci of the primary PTC from the left lobe (T1 to T4), one from the right lobe (T5), and four nodal metastases (M1 to M4). Noncancerous adjacent tissues from the tumor (N1 to N3), the lymph node, and a blood sample were taken as controls. Immediately after surgery, tissues were macro-dissected in several blocks. Each block was cut into two equal parts. One part was embedded in paraffin for histological analyses, and the other part snap-frozen in liquid nitrogen and stored at −80°C. Independent PTC set Twenty-two additional PTCs from Jules Bordet Institute were taken as independent set for hematoxylin and eosin, Ki67, and BRAF staining. They were selected on the basis of sample age and contained 13 BRAF-mutant and 9 BRAF-wild-type (WT; mostly classical PTC variant; Supplemental Table 1). DNA sequencing DNA from tumor and normal thyroid tissues was extracted by using the DNeasy Blood and Tissue Kit (Qiagen, Hilden, Germany) according to the manufacturer’s recommendations. DNA concentrations were spectrophotometrically quantified, and integrity was measured by using an automated gel electrophoresis system (Experion; Bio-Rad, Hercules, CA). Exome DNA sequencing was performed with the Nextera Rapid Capture Expanded exome kit and material was sequenced on the Illumina HiSEquation 2000 platform (Illumina Inc., San Diego, CA) using 100-base paired-end reads, for a yield of ∼180 million reads per sample (analyses in Supplemental Material). IHC IHC staining was performed on 8-µm sections prepared from paraffin-embedded tissues with the primary antibody directed against BRAFV600E (Spring Bioscience, Pleasanton, CA; Clone VE1) or Ki67. The sections were deparaffinized, pretreated with CC1 (EDTA; pH, 8.4), and incubated with the antibody at a 1:100 dilution. The revelation was performed with detection kit ref. 760-501 (Roche, Vilvoorde, Belgium). 3D reconstitution For two of the paraffin tumor blocks, T1 and M1, >100 adjacent slices were cut and stained for BRAFV600E protein. These images were used to reconstruct the associated 3D volumes of the tumor stained for the mutated protein. 3D reconstitution was obtained by using BioVis3D (Montevideo, Uruguay). Images of adjacent slices were manually aligned. Then, on each slice, contours of the stained tumor parts were manually created and were linked, divided, or merged from slice to slice. Finally, BioVis3D rendered a 3D model of these connected 2D contours (Supplemental Video 1). Ki67 quantification For each block, proliferation rates were calculated on scanned images stained for Ki67 by counting the Ki67-stained nuclei and comparing with the nonstained nuclei (23). Ki67 indices were computed separately in multiple regions for the CAFs and follicular parts. This approach provided average proliferation rates and associated estimates of the measurement error across the slice (Supplemental Material; Supplemental Fig. 1). Data availability Sequence data are available in the European Genome-Phenome Archive under accession number EGAB00000001141. All scripts and additional data are available upon request. Study approval The patient gave written informed consent. The medical ethics committee of the Jules Bordet Institute approved this study (The Medical Ethics Committee of Institut Jules Bordet – LEC; 07/06/2012; ref. as-1978). Results Multiregion sequencing and quantitative imaging reveal low cellularity in ∼30% of PTCV600E escaping pathology examination We performed multiregion exome sequencing of a BRAFV600E PTC (Fig. 1A and 1B). The regions included five primary-tumor blocks, four nodal metastases, and three histologically normal thyroid blocks and a normal lymph node. Figure 1. View largeDownload slide Heterogeneity in tumor purity, BRAFV600E clonality, and protein expression in a PTC patient. (A) Sample localization in the patient. (B) Material preparation: each block was halved, one part was frozen for exome sequencing, the other embedded in paraffin [formalin-fixed paraffin embedded (FFPE)] for IHC staining for BRAFV600E and Ki67. (C) Tissue fraction (TF) quantifications (twice the allelic ratios) of BRAFV600E in the DNA of the blocks: Sanger vs next-generation sequencing (NGS). (dotted line: linear fit; grey line: y = x) (D) For each block, BRAFV600E NGS TF vs protein expression ratio computed from IHC images. (E) Same as (D) for Sanger TF against protein expression ratios. (F) For each block, allelic ratios of all somatic mutations (blue) against the three independent measurements of BRAFV600E ratios (in pink: Sanger, NGS, and IHC ratios). (G) Distribution of allelic ratios for BRAFV600E in TCGA PTC with the ratios observed in our samples indicated (purple arrows). M, nodal metastasis; N, normal adjacent thyroid tissue; T, primary tumor block. Figure 1. View largeDownload slide Heterogeneity in tumor purity, BRAFV600E clonality, and protein expression in a PTC patient. (A) Sample localization in the patient. (B) Material preparation: each block was halved, one part was frozen for exome sequencing, the other embedded in paraffin [formalin-fixed paraffin embedded (FFPE)] for IHC staining for BRAFV600E and Ki67. (C) Tissue fraction (TF) quantifications (twice the allelic ratios) of BRAFV600E in the DNA of the blocks: Sanger vs next-generation sequencing (NGS). (dotted line: linear fit; grey line: y = x) (D) For each block, BRAFV600E NGS TF vs protein expression ratio computed from IHC images. (E) Same as (D) for Sanger TF against protein expression ratios. (F) For each block, allelic ratios of all somatic mutations (blue) against the three independent measurements of BRAFV600E ratios (in pink: Sanger, NGS, and IHC ratios). (G) Distribution of allelic ratios for BRAFV600E in TCGA PTC with the ratios observed in our samples indicated (purple arrows). M, nodal metastasis; N, normal adjacent thyroid tissue; T, primary tumor block. The allelic ratio of BRAFV600E estimated from sequencing data ranged from 0% to 30% (Fig. 1C). Quantitative analysis of Sanger sequencing technically validated these allelic ratios (Fig. 1C). Such low ratios could have resulted from low tumor cell fractions, copy number aberration, or subclonal BRAFV600E mutations. Subclonality of the mutation was suggested in previous studies (24, 25). We compared in each region the allelic ratios of BRAFV600E and of all other mutations (Fig. 1F). In all regions but one, BRAFV600E was among the mutations with the largest allelic ratios. This ruled out the subclonality of BRAFV600E in our samples. Paraffin-embedded slices immediately adjacent to each one of the tumor regions used for exome sequencing were stained with an antibody specific of BRAFV600E (Fig. 1A and 1B; Fig. 2A). Taking advantage of the distinctive hue of stained cells, we estimated the areas occupied by BRAFV600E tumor cells and by the entire tumor slice. We then compared the tumor cell area ratio to the tissue fraction estimated from their region-matched BRAFV600E allelic ratio. They were highly consistent (Fig. 1D and 1E), strongly supporting a low ratio of tumor cells within the overall tumor mass. Figure 2. View large Download slide Tumor shape of a PTC in 2D and 3D at histologic resolution. (A) For each tumor sample, a 2D slice stained for BRAFV600E is shown. M1 to M4 are the nodal metastases; M2 to M4 are recurrences. T1 to T5 are the primary tumors. On the top right of each image, a color bar shows the proportion of stained (pink) vs unstained (blue) regions as extracted by image analyses. (B) 3D reconstruction for the T1 primary tumor and the M1 nodal metastasis blocks. Dimensions are given for each part. For T1, one isolated island on the left was found, whereas in M1, all cells belong to the same connected component. In both blocks, the tumor part forms a complex 3D network infiltrating the stroma. See also Supplemental Videos 1 and 2. Figure 2. View large Download slide Tumor shape of a PTC in 2D and 3D at histologic resolution. (A) For each tumor sample, a 2D slice stained for BRAFV600E is shown. M1 to M4 are the nodal metastases; M2 to M4 are recurrences. T1 to T5 are the primary tumors. On the top right of each image, a color bar shows the proportion of stained (pink) vs unstained (blue) regions as extracted by image analyses. (B) 3D reconstruction for the T1 primary tumor and the M1 nodal metastasis blocks. Dimensions are given for each part. For T1, one isolated island on the left was found, whereas in M1, all cells belong to the same connected component. In both blocks, the tumor part forms a complex 3D network infiltrating the stroma. See also Supplemental Videos 1 and 2. Surprisingly, prior pathology review guaranteed a cellularity >70%. As it turned out, the vast fibrotic regions present in the tumor mass were unexpectedly densely cellular. To address how commonly cellularity is underestimated, we obtained allelic ratios for the mutation corresponding to BRAFV600E, as measured from exome sequencing of 211 PTCs from the TCGA (15) (Fig. 1G). Our tumor regions were on the very low end of the TCGA purity range but not below it. Moreover, the a priori pathologic purity threshold for the inclusion of PTC samples in the TCGA was 60%, whereas the purities computed in silico a posteriori were <60% for around one third of the samples (Supplemental Fig. 2). We concluded that tumor cell purity is low in a nontrivial fraction of PTCs and that it is often underestimated during pathology examination. 3D histological-scale reconstruction of low-purity PTCV600E blocks reveals dense connected fractal-like meshes of tumor cells, with an extensive contact surface with the stroma In all regions, BRAFV600E staining revealed isolated patches of tumor cells embedded within an extensive stroma, constituted in the majority by fibroblasts (Fig. 2A). Cells that appear to be isolated in two dimensions, however, could turn out to be part of the same tumor mass when observed in 3D. A protocol was devised to reconstruct and visualize the geometry of the tumor cell mass in a 3D manner (Supplemental Video 1). This 3D reconstruction at histological scale was applied to the nodal metastasis block M1 and the primary tumor block T1 (Fig. 2A and 2B; Supplemental Video 2) of ~1 mm3 each. For block M1, no isolated island of tumor cells was found in the tumor volume reconstructed over a depth of 200 μm (i.e., all cells were part of a single connected component). For block T1, the number of disconnected tumor cell clusters was reduced from a dozen in the 2D slice to an upper bound of two in the 200-μm-deep 3D microvolume. This strongly suggested that the tumor cells within a region formed a single connected component. Interestingly, the 3D volumes departed drastically from any compact ellipsoidal shape. Tumor cells formed a complex volume deeply embedded within the CAFs. Instead of a tumor with an invasive front in contact with the stroma and inner tumor cells in contact mostly with other tumor cells, we observed that most tumor cells were in direct contact with, or nearby, the stroma. CAFs and cancer cells proliferate at similar rates in PTC The results of previous sections raised the question of which cell types from the mixture of stromal and tumor cells drive tumor growth. We performed Ki67 staining on our regions and compared proliferation rates of normal stroma, stroma adjacent to tumor, noncancerous follicular thyroid cells, and tumor cells (Fig. 3A). As expected, proliferation rates were higher in tumor blocks than normal thyroid blocks. Primary tumor regions and metastases had similar proliferation rates. Intriguingly, the proliferation rate of CAFs was not lower than that of follicular thyrocytes. Figure 3. View largeDownload slide Proliferation of epithelial cells and fibroblasts. (Boxplot boxes are delimited by first and third quartiles; the thick segment shows the median; and whiskers extend to the last data points within 1.5 of the box length away from the box.). (A) For each block, a Ki67 proliferation index was derived for cells of thyroid or fibroblast origins in at least one part of the block. The Ki67 ratios are shown on this boxplot for each investigated block in fibroblast (blue) and thyroid (pink) parts. (B) Ki67 indices for an independent series of 22 PTCs, of which 13 were BRAF-mutated. Each point is an average of two Ki67 values for two independent regions on the slides. Figure 3. View largeDownload slide Proliferation of epithelial cells and fibroblasts. (Boxplot boxes are delimited by first and third quartiles; the thick segment shows the median; and whiskers extend to the last data points within 1.5 of the box length away from the box.). (A) For each block, a Ki67 proliferation index was derived for cells of thyroid or fibroblast origins in at least one part of the block. The Ki67 ratios are shown on this boxplot for each investigated block in fibroblast (blue) and thyroid (pink) parts. (B) Ki67 indices for an independent series of 22 PTCs, of which 13 were BRAF-mutated. Each point is an average of two Ki67 values for two independent regions on the slides. To generalize this finding, we measured Ki67 indices in the stromal and tumor compartments of 22 independent PTC, including 13 BRAFV600E and 9 BRAFWT tumors (Fig. 3B). Importantly, these samples were not selected for specific tumor cell purity. For each tumor, one slice was stained for Ki67. The ratio of Ki67-stained nuclei for tumor cells was not higher than for CAFs, suggesting that the density of cycling cells in the fibrous stroma was high. BRAFV600E is associated with a desmoplastic phenotype, high proliferation rate, and dedifferentiation We combined image, genomic, and transcriptomic analyses of the 496 PTCs from the TCGA (15) to further characterize the phenotype of PTCV600E and how it relates to the relative sizes of the tumor and stromal compartments. In the TCGA cohort, we downloaded the PTC histological hematoxylin and eosin–stained slide images from the TCGA cancer digital archives (http://cancer.digitalslidearchive.net). Because fibrosis presented a distinctive color hue (Fig. 4A), it was possible to automatically estimate fibrotic content as the ratio of the fibrotic areas to the total tumor area. Fibrotic content was higher in PTCV600E than in other PTCs (Fig. 4B). RAS-mutated tumors had the least fibrous tissue. Figure 4. View largeDownload slide PTCV600E gene expression, purity, and fibrosis phenotypes in TCGA PTCs. (A) Top: histological image of one PTC from TCGA, where fibrosis appears with a pink hue distinct from the blue hue of surrounding tissues (hematoxylin and eosin staining, scanned at ×40, unzoomed, cropped). Bottom: selection of fibrosis (purple) and surrounding tissues (black) based on their respective hues. The fibrosis content was calculated from these two areas. (B) Fibrosis content by mutational group for all TCGA PTC that had histological images available (n = 331). (C) Purity by mutational group of TCGA PTC as computed by ABSOLUTE method in the original publication (n = 364) (15). (D) Fibrosis by mutational group after correction for purity of the samples using a robust linear regression model (N = 331). (E) Projection of 423 expression profiles and gene signatures (arrows) onto the first two components computed from TCGA PTC (circles) and adjacent normal tissues (triangles). Gene expression signatures are median values of genes related to thyroid, T cells, and fibroblasts, proliferation, and BRAF-RAS genes. Arrows point to the directions of higher expression of the genes present in the underlying signatures. Figure 4. View largeDownload slide PTCV600E gene expression, purity, and fibrosis phenotypes in TCGA PTCs. (A) Top: histological image of one PTC from TCGA, where fibrosis appears with a pink hue distinct from the blue hue of surrounding tissues (hematoxylin and eosin staining, scanned at ×40, unzoomed, cropped). Bottom: selection of fibrosis (purple) and surrounding tissues (black) based on their respective hues. The fibrosis content was calculated from these two areas. (B) Fibrosis content by mutational group for all TCGA PTC that had histological images available (n = 331). (C) Purity by mutational group of TCGA PTC as computed by ABSOLUTE method in the original publication (n = 364) (15). (D) Fibrosis by mutational group after correction for purity of the samples using a robust linear regression model (N = 331). (E) Projection of 423 expression profiles and gene signatures (arrows) onto the first two components computed from TCGA PTC (circles) and adjacent normal tissues (triangles). Gene expression signatures are median values of genes related to thyroid, T cells, and fibroblasts, proliferation, and BRAF-RAS genes. Arrows point to the directions of higher expression of the genes present in the underlying signatures. The area of the fibrotic regions does not necessarily reflect the number of cells within these regions. We therefore investigated tumor purity of TCGA PTC (15), as measured in silico by the ABSOLUTE method (26). Purity was lower in BRAFV600E PTCs than in BRAFWT PTCs and especially higher in RAS-mutated PTCs (Fig. 4C). To further define the activity of these fibrotic regions, we corrected the fibrosis ratios for the purity of the samples by using a robust linear model (Fig. 4D). The association of BRAFV600E with fibrosis remained highly significant. Taken together with the Ki67 data from our cohort, these results suggest that PTCV600E is associated with a desmoplastic phenotype: These tumors include more fibroblasts, which proliferate more and produce more fibrosis. To gain a broader view of the phenotype of individual PTCs from the TCGA, we derived five gene expression signatures capturing different aspects of PTC biology. Two were derived by following the procedure of the original TCGA study, including genes associated with thyroid differentiation and with the BRAF-RAS axis (15). We also derived three additional signatures for fibroblasts, T cells, and proliferation. Associations between BRAF mutational status and each one of the five gene expression signatures were statistically significant (Mann-Whitney tests; false discovery rates <10−5). To visualize the global trends in gene expression, we performed a principal component analysis of TCGA RNA-seq data. Expression profiles were projected on the two first components, together with the five signature vectors. On Fig. 4F, points and triangles depict transcriptomes of PTC and normal thyroid tissues, respectively. Their relative projection on each signature vector is closely related to the relative average expression of the genes making up the underlying gene signatures. PTCV600E grouped separately from normal tissues along axes of thyroid differentiation, proliferation, and BRAF-RAS expression differences. As expected from the TCGA study, they further separated from BRAFWT PTCs along the vector representing thyroid differentiation and BRAF-RAS expression differences, but also along the vectors standing for fibroblasts, lymphocytes, and proliferation (Fig. 4F). Interestingly, the fibroblasts and proliferation axes were strongly associated in these PTCs, which was reflected in their almost identical directions on Fig. 4F. Together the higher proliferation and dedifferentiation indices of PTCV600E point to a more aggressive phenotype. Discussion Textbooks and medical imaging suggest that cancers expand as a compact near-ellipsoid ball with an inner core and an invasive front in contact with noncancerous tissues. At histological scale, the 3D reconstruction of our low-purity PTC revealed a morphology departing radically from this archetype. Far from an ellipsoid, this morphology would be better depicted as a branching fractal-like aggregate of cancer cells deeply embedded within the stroma. In this morphology, the concepts of inner core and invasive front break down because all tumor cells are within short distance from the stroma. The 3D reconstructions also suggested that tumor cells belonged to a single connected component within the ∼1-mm3 volumes investigated. Thus, all cancer cells were in contact with or close to both stromal and cancer cells. CAFs lead the collective invasion of carcinoma cells (4). The same phenomenon could be active in PTC, as elongating patterns of neoplastic cells into nonneoplastic cells were described in microPTC (7). Transforming growth factor-β (TGF-β) signaling was involved in many tumor-stroma interactions involving EMT, proliferation, and activating signals. PTC cells could send signals to the surrounding stroma, potentially involving TGF-β1 (27), with increased response or signal in PTCV600E. Our analysis also argues against the textbook view that tumor expansion is driven by cancer cell proliferation alone. Cancer cells represented a small fraction of the total number of cells making the tumor mass, <10% in several regions. In such circumstances, we ask how the tumor still reaches a clinically detectable volume and whether excess proliferation is an exclusive property of the tumor cells. We measured proliferation rates in the tumor and stromal compartments of our original low-purity regions and associated normal-thyroid tissues and in an independent series of PTCs. In contrast to the former, the latter series was not selected for specific purity or fibrotic content. Proliferation was higher in tumor than in normal tissues for both compartments. Our estimations based on Ki67 staining suggested that CAFs proliferate faster than tumor cells (Fig. 3B). However, unstained nuclei are more easily detected in cancer cells than in fibroblasts, which may bias any direct comparison of proliferation between these compartments. At this point, we can assert only that the proliferation of fibroblasts is increased in PTCs compared with normal thyroid. Further research is needed to quantify precisely the relative contribution of both compartments to tumor expansion. Pathology examination initially suggested that our samples had purities >70%. Subsequent sequencing and computational analyses demonstrated that they were much lower. Fibrotic regions, it turned out, were more cellular than expected. Similarly, inclusion of samples in TCGA thyroid cancer dataset required a purity >60% upon pathology review, but subsequent sequencing and analysis demonstrated that one third of the samples did not meet this criterion. Purity could be underestimated in other studies. Our samples were at the lower end of the purity spectrum, but they were not unique: 3.5% of the 342 TCGA samples analyzed had purities <25%. Given that the inclusion criteria of the TCGA introduced a selection bias toward higher purity, 25% would be an upper-bound estimate of the true purity range of PTC. Incidentally, the association of low purity with BRAFV600E implies that BRAFV600E incidence could also be underestimated in TCGA and possibly other studies. This observation generalizes to any confounding genetic factor correlated with purity. To interpret the genome-derived purities, we also analyzed the fibrotic content of TCGA PTC from tumor images and the expression of fibroblast-associated genes. We observed a wide spectrum of purity, fibrosis, and fibroblast gene expression scores. Within this spectrum, nearly all RAS-mutated PTCs had highest purity, lower fibrosis, and lower fibroblast gene expression. By contrast, other tumors, most of them harboring BRAFV600E, covered the entire range of scores, including a sizable fraction of tumors with low purity, high fibrosis, and high fibroblast gene expression. Previous literature pointed to a role for CAF in PTC. In one study, PTCV600E showed more fibrosis/desmoplasia and infiltrative growth (28). In another study, microPTCV600E also had more fibrosis content (7). Interestingly, both the image-derived fibrotic and the fibroblast gene expression scores were associated with BRAFV600E independently of purity in our analyses. This suggests that the mutational status of PTC is associated with fibrosis but also the density of CAF and their activation state. We showed that PTCV600E have a higher expression of proliferation-associated genes and of the Ki67 protein and lower expression of thyroid-differentiation genes. Potential links have been described between aggressiveness and tumor-stroma crosstalk in PTC (29). For example, PTCV600E could be more invasive because of decreased CDH1 expression (30) and more susceptible to TGF-β–induced EMT through an MAPK-dependent process (31). Furthermore, the cross-talk involving TGF-β at the invasive front led to more aggressive and refractory PTCV600E in PCCL3 cells (32). CAFs and endothelial cells may also express EMT markers (33). Given the high fibroblast content of PTCV600E revealed here, expression of EMT markers in bulk tissue profiling should be interpreted with caution. Altogether, our results depict a more aggressive phenotype of PTCV600E, aided by extensive, more active and proliferative CAF (i.e., a higher desmoplastic reaction associated with BRAFV600E in PTCs). These tumors could in theory be treated accordingly. Approved drugs targeting the BRAF-mutated pathway would represent good candidates. Although they are still at an early stage of clinical assessment in thyroid cancer, PTCs might respond to a lesser degree to BRAF inhibitors compared with melanoma and other cancer types with MAPK pathway activation (34). Interestingly, higher CAF content and a stiffer extracellular matrix protect melanomaV600E cells against BRAF-pathway inhibitors (35). The higher density and activation of CAFs in PTCs uncovered here may similarly induce resistance (35). Hence, our results would be compatible with improved treatment response via combination of ECM and BRAF inhibitors in PTCV600E (35). Abbreviations: Abbreviations: 2D two-dimensional 3D three-dimensional CAF cancer-associated fibroblast EMT epithelial-to-mesenchymal transition IHC immunohistochemistry MAPK mitogen-activated protein kinase PTC papillary thyroid carcinoma TCGA The Cancer Genome Atlas TGF-β transforming growth factor-β WT wild type Acknowledgments Part of the results presented here is based on data generated by the TCGA Research Network (http://cancergenome.nih.gov/). The authors thank Chantale Degraef for her excellent technical assistance and Jonas Demeulemeester for proofreading and helpful advice and discussion. Financial Support: M.T. was supported by a Fonds pour la Formation à la Recherche dans l’Industrie et dans l’Agriculture grant. This work was supported by the Fonds de la Recherche Scientifique FNRS-FRSM (C.M.), Walloon Excellence in Life Sciences and Biotechnology, Plan Cancer Belgique, les Amis de l’Institut Bordet (C.M.), and Fondation Van Buuren (C.M.). V.D. was funded by Fonds De La Recherche Scientifique, grant J009714F. Author Contributions: M.T. and D.G. conducted bioinformatic analyses. M.T. performed image analyses. A.A. processed the PTC sample for next-generation sequencing and performed Sanger sequencing experiments. A.A. and M.T. reconstructed the 3D microvolumes. M.T., A.A., C.M., and V.D. interpreted the results. M.T., J.E.D., C.M., and V.D. wrote the manuscript. S.L.P., C.M., and V.D. designed the experiments. G.A. collected patient samples. N.d.S.A., L.C., and D.L. collected and reviewed the histopathological slices of the independent PTC dataset and the sequenced PTC case. A.A., T.C., and I.L. performed immune stainings. Disclosure Summary: The authors have nothing to disclose. References 1. Waclaw B, Bozic I, Pittman ME, Hruban RH, Vogelstein B, Nowak MA. A spatial model predicts that dispersal and cell turnover limit intratumour heterogeneity. Nature . 2015; 525( 7568): 261– 264. Google Scholar CrossRef Search ADS PubMed 2. Sharma M, Sah P, Sharma SS, Radhakrishnan R. Molecular changes in invasive front of oral cancer. J Oral Maxillofac Pathol . 2013; 17( 2): 240– 247. Google Scholar CrossRef Search ADS PubMed 3. Friedl P, Alexander S. Cancer invasion and the microenvironment: plasticity and reciprocity. Cell . 2011; 147( 5): 992– 1009. Google Scholar CrossRef Search ADS PubMed 4. Gaggioli C, Hooper S, Hidalgo-Carcedo C, Grosse R, Marshall JF, Harrington K, Sahai E. Fibroblast-led collective invasion of carcinoma cells with differing roles for RhoGTPases in leading and following cells. Nat Cell Biol . 2007; 9( 12): 1392– 1400. Google Scholar CrossRef Search ADS PubMed 5. Eloy C, Santos J, Soares P, Sobrinho-Simões M. The preeminence of growth pattern and invasiveness and the limited influence of BRAF and RAS mutations in the occurrence of papillary thyroid carcinoma lymph node metastases. Virchows Arch . 2011; 459( 3): 265– 276. Google Scholar CrossRef Search ADS PubMed 6. Liu Z, Kakudo K, Bai Y, Li Y, Ozaki T, Miyauchi A, Taniguchi E, Mori I. Loss of cellular polarity/cohesiveness in the invasive front of papillary thyroid carcinoma, a novel predictor for lymph node metastasis; possible morphological indicator of epithelial mesenchymal transition. J Clin Pathol . 2011; 64( 4): 325– 329. Google Scholar CrossRef Search ADS PubMed 7. Tallini G, de Biase D, Durante C, Acquaviva G, Bisceglia M, Bruno R, Bacchi Reggiani ML, Casadei GP, Costante G, Cremonini N, Lamartina L, Meringolo D, Nardi F, Pession A, Rhoden KJ, Ronga G, Torlontano M, Verrienti A, Visani M, Filetti S. BRAF V600E and risk stratification of thyroid microcarcinoma: a multicenter pathological and clinical study. Mod Pathol . 2015; 28( 10): 1343– 1359. Google Scholar CrossRef Search ADS PubMed 8. Lambin P, Rios-Velazquez E, Leijenaar R, Carvalho S, van Stiphout RGPM, Granton P, Zegers CML, Gillies R, Boellard R, Dekker A, Aerts HJWL. Radiomics: extracting more information from medical images using advanced feature analysis. Eur J Cancer . 2012; 48( 4): 441– 446. Google Scholar CrossRef Search ADS PubMed 9. Vasko V, Espinosa AV, Scouten W, He H, Auer H, Liyanarachchi S, Larin A, Savchenko V, Francis GL, de la Chapelle A, Saji M, Ringel MD. Gene expression and functional evidence of epithelial-to-mesenchymal transition in papillary thyroid carcinoma invasion. Proc Natl Acad Sci USA . 2007; 104( 8): 2803– 2808. Google Scholar CrossRef Search ADS PubMed 10. Kirkby MJ. Review of The Fractal Geometry of Nature by Benoit B. Mandelbrot. Earth Surf Process Landf . 1983; 8( 4): 406– 406. Google Scholar CrossRef Search ADS 11. Tambasco M, Magliocco AM. Relationship between tumor grade and computed architectural complexity in breast cancer specimens. Hum Pathol . 2008; 39( 5): 740– 746. Google Scholar CrossRef Search ADS PubMed 12. Mitrovic B, Schaeffer DF, Riddell RH, Kirsch R. Tumor budding in colorectal carcinoma: time to take notice. Mod Pathol . 2012; 25( 10): 1315– 1325. Google Scholar CrossRef Search ADS PubMed 13. Kondo T, Ezzat S, Asa SL. Pathogenetic mechanisms in thyroid follicular-cell neoplasia. Nat Rev Cancer . 2006; 6( 4): 292– 306. Google Scholar CrossRef Search ADS PubMed 14. Rahib L, Smith BD, Aizenberg R, Rosenzweig AB, Fleshman JM, Matrisian LM. Projecting cancer incidence and deaths to 2030: the unexpected burden of thyroid, liver, and pancreas cancers in the United States. Cancer Res . 2014; 74( 11): 2913– 2921. Google Scholar CrossRef Search ADS PubMed 15. Cancer Genome Atlas Research Network. Integrated genomic characterization of papillary thyroid carcinoma. Cell . 2014; 159( 3): 676– 690. CrossRef Search ADS PubMed 16. Roger PP, van Staveren WCG, Coulonval K, Dumont JE, Maenhaut C. Signal transduction in the human thyrocyte and its perversion in thyroid tumors. Mol Cell Endocrinol . 2010; 321( 1): 3– 19. Google Scholar CrossRef Search ADS PubMed 17. Ghossein RA, Katabi N, Fagin JA. Immunohistochemical detection of mutated BRAF V600E supports the clonal origin of BRAF-induced thyroid cancers along the spectrum of disease progression. J Clin Endocrinol Metab . 2013; 98( 8): E1414– E1421. Google Scholar CrossRef Search ADS PubMed 18. Kim SJ, Lee KE, Myong JP, Park JH, Jeon YK, Min HS, Park SY, Jung KC, Koo H, Youn Y-K. BRAF V600E mutation is associated with tumor aggressiveness in papillary thyroid cancer. World J Surg . 2012; 36( 2): 310– 317. Google Scholar CrossRef Search ADS PubMed 19. Sarne DH. A piece of the puzzle: what does BRAF status mean in the management of patients with papillary thyroid carcinoma? J Clin Endocrinol Metab . 2012; 97( 9): 3094– 3096. Google Scholar CrossRef Search ADS PubMed 20. Jing FJ, Liang J, Liang ZY, Meng C, Long W, Li XY, Lin YS. BRAF(V600E) mutation is not a positive predictor for distant metastasis in sporadic papillary thyroid carcinoma. Chin Med J (Engl) . 2013; 126( 16): 3013– 3018. Google Scholar PubMed 21. Sancisi V, Nicoli D, Ragazzi M, Piana S, Ciarrocchi A. BRAFV600E mutation does not mean distant metastasis in thyroid papillary carcinomas. J Clin Endocrinol Metab . 2012; 97( 9): E1745– E1749. Google Scholar CrossRef Search ADS PubMed 22. Nikiforova MN, Kimura ET, Gandhi M, Biddinger PW, Knauf JA, Basolo F, Zhu Z, Giannini R, Salvatore G, Fusco A, Santoro M, Fagin JA, Nikiforov YE. BRAF mutations in thyroid tumors are restricted to papillary carcinomas and anaplastic or poorly differentiated carcinomas arising from papillary carcinomas. J Clin Endocrinol Metab . 2003; 88( 11): 5399– 5404. Google Scholar CrossRef Search ADS PubMed 23. Reid MD, Bagci P, Ohike N, Saka B, Erbarut Seven I, Dursun N, Balci S, Gucer H, Jang K-T, Tajiri T, Basturk O, Kong SY, Goodman M, Akkas G, Adsay V. Calculation of the Ki67 index in pancreatic neuroendocrine tumors: a comparative analysis of four counting methodologies. 2015; 28( 5): 686– 694. 24. Guerra A, Sapio MR, Marotta V, Campanile E, Rossi S, Forno I, Fugazzola L, Budillon A, Moccia T, Fenzi G, Vitale M. The primary occurrence of BRAF(V600E) is a rare clonal event in papillary thyroid carcinoma. J Clin Endocrinol Metab . 2012; 97( 2): 517– 524. Google Scholar CrossRef Search ADS PubMed 25. de Biase D, Cesari V, Visani M, Casadei GP, Cremonini N, Gandolfi G, Sancisi V, Ragazzi M, Pession A, Ciarrocchi A, Tallini G. High-sensitivity BRAF mutation analysis: BRAF V600E is acquired early during tumor development but is heterogeneously distributed in a subset of papillary thyroid carcinomas. J Clin Endocrinol Metab . 2014; 99( 8): E1530– E1538. Google Scholar CrossRef Search ADS PubMed 26. Carter SL, Cibulskis K, Helman E, McKenna A, Shen H, Zack T, Laird PW, Onofrio RC, Winckler W, Weir BA, Beroukhim R, Pellman D, Levine DA, Lander ES, Meyerson M, Getz G. Absolute quantification of somatic DNA alterations in human cancer. Nat Biotechnol . 2012; 30( 5): 413– 421. Google Scholar CrossRef Search ADS PubMed 27. Zhang J, Wang Y, Li D, Jing S. Notch and TGF-β/Smad3 pathways are involved in the interaction between cancer cells and cancer-associated fibroblasts in papillary thyroid carcinoma. Tumour Biol . 2014; 35( 1): 379– 385. Google Scholar CrossRef Search ADS PubMed 28. Finkelstein A, Levy GH, Hui P, Prasad A, Virk R, Chhieng DC, Carling T, Roman SA, Sosa JA, Udelsman R, Theoharis CG, Prasad ML. Papillary thyroid carcinomas with and without BRAF V600E mutations are morphologically distinct. Histopathology . 2012; 60( 7): 1052– 1059. Google Scholar CrossRef Search ADS PubMed 29. Nucera C, Lawler J, Parangi S. BRAF(V600E) and microenvironment in thyroid cancer: a functional link to drive cancer progression. Cancer Res . 2011; 71( 7): 2417– 2422. Google Scholar CrossRef Search ADS PubMed 30. Baquero P, Sánchez-Hernández I, Jiménez-Mora E, Orgaz JL, Jiménez B, Chiloeches A. (V600E)BRAF promotes invasiveness of thyroid cancer cells by decreasing E-cadherin expression through a Snail-dependent mechanism. Cancer Lett . 2013; 335( 1): 232– 241. Google Scholar CrossRef Search ADS PubMed 31. Knauf JA, Sartor MA, Medvedovic M, Lundsmith E, Ryder M, Salzano M, Nikiforov YE, Giordano TJ, Ghossein RA, Fagin JA. Progression of BRAF-induced thyroid cancer is associated with epithelial-mesenchymal transition requiring concomitant MAP kinase and TGFβ signaling. Oncogene . 2011; 30( 28): 3153– 3162. Google Scholar CrossRef Search ADS PubMed 32. Riesco-Eizaguirre G, Rodríguez I, De la Vieja A, Costamagna E, Carrasco N, Nistal M, Santisteban P. The BRAFV600E oncogene induces transforming growth factor β secretion leading to sodium iodide symporter repression and increased malignancy in thyroid cancer. Cancer Res . 2009; 69( 21): 8317– 8325. Google Scholar CrossRef Search ADS PubMed 33. Buehler D, Hardin H, Shan W, Montemayor-Garcia C, Rush PS, Asioli S, Chen H, Lloyd RV. Expression of epithelial-mesenchymal transition regulators SNAI2 and TWIST1 in thyroid carcinomas. Mod Pathol . 2013; 26( 1): 54– 61. Google Scholar CrossRef Search ADS PubMed 34. Hyman DM, Puzanov I, Subbiah V, Faris JE, Chau I, Blay J-Y, Wolf J, Raje NS, Diamond EL, Hollebecque A, Gervais R, Elez-Fernandez ME, Italiano A, Hofheinz R-D, Hidalgo M, Chan E, Schuler M, Lasserre SF, Makrutzki M, Sirzen F, Veronese ML, Tabernero J, Baselga J. Vemurafenib in Multiple Nonmelanoma Cancers with BRAF V600 Mutations. N Engl J Med . 2015; 373( 8): 726– 736. Google Scholar CrossRef Search ADS PubMed 35. Hirata E, Girotti MR, Viros A, Hooper S, Spencer-Dene B, Matsuda M, Larkin J, Marais R, Sahai E. Intravital imaging reveals how BRAF inhibition generates drug-tolerant microenvironments with high integrin β1/FAK signaling. Cancer Cell . 2015; 27( 4): 574– 588. Google Scholar CrossRef Search ADS PubMed Copyright © 2018 Endocrine Society
Journal of Clinical Endocrinology and Metabolism – Oxford University Press
Published: Mar 1, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera