The cell of origin in differentiated thyroid cancer – the follicular cell – gives rise to two main tumor types: follicular thyroid carcinoma (FTC) and papillary thyroid carcinoma (PTC), both of which are overrepresented in women (Dralle et al., 2015; Williams, 2015). As indicated by the nomenclature, FTC and PTC possess different features of tumor growth and differentiation that influence clinical outcome; they are, therefore, considered as separate cancer entities (Dralle et al., 2015). The fact that mutation of different members of the small GTPase family RAS – mainly of NRAS and BRAF – predominantly associate with FTC and PTC, respectively (Howell et al., 2013; Xing, 2005), suggests that mutation identity can influence the morphogenesis of a distinct carcinoma phenotype. However, these driver mutations occur, although with varying frequencies, in nearly all thyroid cancer types. For example, PTC induced by BRAF carrying a Val600 to Glu (V600E) point mutation (BRAFV600E) comprises several subtypes including the follicular variant (Afkhami et al., 2016). A single PTC tumor may also display a mixture of growth patterns, the predominant one being decisive for diagnostic subtyping. Most PTCs have a low somatic mutation burden, indicating that genomic instability is not a critical factor except in advanced tumor stages (Cancer Genome Atlas Research Network, 2014). Although transcriptional profiling distinguishes between PTCs as being ‘RAS-like’ and ‘BRAF-like’ neoplasms with different levels of tumor dedifferentiation and aggressiveness (Cancer Genome Atlas Research Network, 2014), the underlying mechanisms of the morphogenetic traits that give rise to heterogeneous tumor phenotypes in the thyroid are unknown.

Tumor heterogeneity conceptually implies diversification of cancer cell properties – genetically, morphologically and biochemically – that are involved in tumor progression. According to the clonal evolution model (Swanton, 2012), tumors arise from a single mutated cell that, upon accumulation of additional somatic mutations, gives rise to a tumor clone that possesses a growth advantage. Further branched evolution of heterogeneous subclones contributes to tumor progression and provides a selection mechanism to escape anti-cancer drug treatment (Turajlic et al., 2019). An alternative, less-recognized mechanism concerns the possible involvement of a multiclonal tumor origin (Parsons, 2018), which implies that two or more independent clones cooperate in tumor development and heterogeneous tumor growth. Such clonal cooperation might be necessary for mutant cells to resist competition with non-mutant cells. Escaping surveillance mechanisms associated with normal tissue homeostasis (Bowling et al., 2019) may not only be decisive for tumor development but also influence later stages of cancer progression (Cheung et al., 2016; Echeverria et al., 2018; Reeves et al., 2018). In thyroid cancer patients, mutation analysis infers that PTC, the most common type of thyroid cancer, is a strictly monoclonal tumor (Cancer Genome Atlas Research Network, 2014). However, as recently reviewed (Fugazzola et al., 2020), there are several divergent reports of genetically heterogeneous tumor cell populations that argue against the prevailing concept of monoclonality, suggesting that PTC development comprises subclonal or even oligoclonal events. To date, there are no experimental studies that address the clonal origin of differentiated thyroid cancer and the role of clonality in thyroid tumor development.

Mouse models have been invaluable in studies of thyroid cancer progression and the evaluation of targeted therapies (Landa and Knauf, 2019). However, since conditional expression of oncogenes, such as mutant BRAF, encompasses the majority of cells, there are no current models that reliably replicate tumor initiation and early events of the carcinogenic process within a preserved thyroid tissue microenvironment. Notably, loss of thyroid function, accompanying synchronous activation of BrafCA that encodes the BRAFV600E oncoprotein, rapidly generates pronounced goitre growth and leads to global disruption of the normal follicular organization (Chakravarty et al., 2011), which invalidates the monitoring of any discrete cellular changes presumed to characterize focal tumorigenesis. To overcome these obstacles, we have adopted a novel tumorigenic approach based on spontaneous Cre-mediated recombination, which occurs at significant levels under non-induced conditions in TgCreERT2;BrafCA/+ mice that conditionally express BrafV600E within the thyroid (Charles et al., 2011). This enabled us to investigate through lineage tracing the earliest stages of BrafV600E-induced tumor development and to elucidate the clonal origin of tumor heterogeneity. Our results indicate that tissue organization designated by follicle heterogeneity delimits the effective cancerization field and, hence, determines the fate of thyroid cells subjected to oncogene activation. Postnatal nascent follicles are particularly susceptible to Braf mutations, as they are prone to develop oligoclonal lesions that escape competition and give rise to PTCs of diverse phenotypes in mice.

We obtained our sporadic thyroid cancer mouse model TgCreERT2;BrafCA/+ by crossing the established BrafCA and TgCreERT2 mouse lines (Dankort et al., 2007; Undeutsch et al., 2014) to conditionally express BRAFV600E in the thyroid under control of the thyroglobulin (Tg) promoter, as previously reported (Charles et al., 2011). To yield our current TgCreERT2;BrafCA/+ models, mutant mice were not induced by tamoxifen, in order to elucidate whether spontaneous activation of mutant BRAF due to leaky activity of the bacteriophagal Cre recombinase reproduce sporadic thyroid cancer development. In situ gland volume measurements showed that the thyroid gradually increased in size, with growth accelerating between 6 and 12 months of age in mutant animals (Fig. 1A,B). Notably, at 6 months, i.e. before great variations in thyroid size became evident, the gland was significantly larger in females than in male mutants (Fig. 1C). Consistent with the stochastic generation of tumors with different growth properties, the relative enlargement of the left and right thyroid lobes differed increasingly with age in the majority of mice (Fig. 1D and E). Interestingly, 83% of 12-months-old mutants (n=18) showed a higher left-to-right lobe ratio (Fig. 1D). The fact that larger tumors were predominant in the left lobe was reinforced by a trend of the opposite-lobe asymmetry in aging control animals (Fig. 1D). A single-sided preference is evident for thyroid developmental defects in both mice (Manley and Capecchi, 1995) and humans (Shabana et al., 2000), but has not previously been reported for neoplastic lesions. As documented by magnetic resonance imaging (MRI), heterogeneous tumor growth explained the lobe size differences (Fig. 1G,H; Movie 1). Mouse MRI also revealed that cystic tumors were predominantly located at the periphery of the lobes, whereas more solid tumor portions often took a central or medial location in the lobes.

Fig. 1.

Occurrence of papillary thyroid carcinoma (PTC) in non-induced TgCreERT2;BrafCA/+ mice devoid of tamoxifen injections. Data are from wild-type (wt) and BrafCA mutant mice at age 3-18 months (mo). Thyroid volumes were estimated from lobe diameter measurements. (A,E) Shown are in situ micrographs of frontal views of enlarged thyroids. (B) Thyroid volumes plotted over time. (C) Thyroid volume in relation to the sex plotted over time. (D) Ratio of left to right lobe – i.e. asymmetric lobe growth – plotted over time. Mean±s.d; *P<0.005; **P<0.0001. For B and D, numbers of mice were wt (n=12) and mutant (n=17) at 3 months; wt (n=16) and mutants (n=20) at 6 months; wt (n=12) and mutants (n=18) at 12 months. (F) Immunostaining for cytokeratin 19 (CK19) showing increased protein levels, consistent with raised CK19 levels observed in human PTCs. (G,H) T2-weighted MRI image (G) of the same thyroid specimen as shown in E and F (for entire stack series, see Movie 1) and apparent diffusion coefficient (ADC) color map (H) of the same image. The color bar relates to solid (red) and cystic (yellow) tumors based on the ADC (µm2/ms). cy, corresponding cystic tumor portions. Used technology delimited resolution of images. (I-K) H&E staining showing inter- and intra-tumor heterogeneity of multifocal PTCs. Images of additional tumors present in the same thyroid specimen are shown in Figs S1 and S2. Two adjacent PTC tumor foci are encircled (1 and 2) and shown in I; the boxed area in circle 1 is shown magnified in I′, indicating a transition of the tumor growth pattern. Immunostaining for NKX2-1 indicating its downregulation in the solid tumor portion, is shown in J; the parallel section of J is shown in panel l′. Nuclear characteristics of tumor cells are shown in K – magnification of a region within the first PTC tumor (circle 1) from I; the interface between lumen and stroma of the tumor tissue is indicted by dashed lines. Immunostaining for NKX2-1 (L) within the second PTC tumor shown in I (circle 2). The boxed area in L is shown magnified in L′. (M) Sketch outlining the papillary tumor growth as shown in L′. Anatomical orientation: D, dorsal; L, left; R, right; V, ventral. cl, classic variant of PTC; so, solid variant of PTC; t, trachea. Arrows indicate the tumor stalk; arrowheads indicate the follicular wall enclosing the tumor. Scale bars: 500 µm (F,I), 100 µm (I′,J,L), 50 µm (K).

Fig. 1.

Occurrence of papillary thyroid carcinoma (PTC) in non-induced TgCreERT2;BrafCA/+ mice devoid of tamoxifen injections. Data are from wild-type (wt) and BrafCA mutant mice at age 3-18 months (mo). Thyroid volumes were estimated from lobe diameter measurements. (A,E) Shown are in situ micrographs of frontal views of enlarged thyroids. (B) Thyroid volumes plotted over time. (C) Thyroid volume in relation to the sex plotted over time. (D) Ratio of left to right lobe – i.e. asymmetric lobe growth – plotted over time. Mean±s.d; *P<0.005; **P<0.0001. For B and D, numbers of mice were wt (n=12) and mutant (n=17) at 3 months; wt (n=16) and mutants (n=20) at 6 months; wt (n=12) and mutants (n=18) at 12 months. (F) Immunostaining for cytokeratin 19 (CK19) showing increased protein levels, consistent with raised CK19 levels observed in human PTCs. (G,H) T2-weighted MRI image (G) of the same thyroid specimen as shown in E and F (for entire stack series, see Movie 1) and apparent diffusion coefficient (ADC) color map (H) of the same image. The color bar relates to solid (red) and cystic (yellow) tumors based on the ADC (µm2/ms). cy, corresponding cystic tumor portions. Used technology delimited resolution of images. (I-K) H&E staining showing inter- and intra-tumor heterogeneity of multifocal PTCs. Images of additional tumors present in the same thyroid specimen are shown in Figs S1 and S2. Two adjacent PTC tumor foci are encircled (1 and 2) and shown in I; the boxed area in circle 1 is shown magnified in I′, indicating a transition of the tumor growth pattern. Immunostaining for NKX2-1 indicating its downregulation in the solid tumor portion, is shown in J; the parallel section of J is shown in panel l′. Nuclear characteristics of tumor cells are shown in K – magnification of a region within the first PTC tumor (circle 1) from I; the interface between lumen and stroma of the tumor tissue is indicted by dashed lines. Immunostaining for NKX2-1 (L) within the second PTC tumor shown in I (circle 2). The boxed area in L is shown magnified in L′. (M) Sketch outlining the papillary tumor growth as shown in L′. Anatomical orientation: D, dorsal; L, left; R, right; V, ventral. cl, classic variant of PTC; so, solid variant of PTC; t, trachea. Arrows indicate the tumor stalk; arrowheads indicate the follicular wall enclosing the tumor. Scale bars: 500 µm (F,I), 100 µm (I′,J,L), 50 µm (K).

Advanced tumor stages were further analyzed in serial-sectioned thyroids from non-induced TgCreERT2;BrafCA/+ mice between 12 and 22 months (n=15). Consistent with PTC features in humans (Rorive et al., 2002), we observed increased levels of cytokeratin 19 (KRT19, hereafter referred to as CK19) in tumor cells of mutant mice (Fig. 1F; Fig. S1A,B). Moreover, reminiscent of PTC subtypes classified histologically, the tumors showed a highly variable growth pattern comprising (i) classic or conventional PTC with an abundance of papillary formations (Fig. 1I-M; Fig. S1B,C) and the characteristic ground glass nuclear features of the tumor cells (Fig. 1K); (ii) cystic PTC with hobnail-like features of the epithelial lining (Fig. S1E); (iii) solid variant PTC essentially devoid of papillary growth (Figs. S1G, S2A-A′) and; (iv) tall-cell variant PTC with an unusual cylindrical tumor cell shape (Fig. S2D,D′). Altered expression levels of the key thyroid transcription factor NKX2-1 (Fernández et al., 2015) were evident among tumors and also within a single tumor (Fig. 1J,L,L′; Fig. S2B,B′), presumably reflecting heterogeneous tumor properties. Notably, tumor cells invaded stroma-rich tissue either by collective migration with a preserved ability to form follicles (Figs S1D, S2D,D′) or by undergoing partial epithelial-mesenchymal transition (EMT) characterized by diminished expression and disrupted localization of E-cadherin (Figs S1G, S2B,B′ and S2C,D). Occasionally, tumor cells infiltrated extra-thyroidal tissues (Fig. S2D,D′).

Altogether, these findings indicated that the sporadic activation of BrafCA in mice generates multifocal thyroid carcinomas with distinctive growth patterns that mimic PTC subtypes in humans. Notably, the penetrance of Braf mutation was 100%, although with highly variable tumor sizes and phenotypes occurring among individuals as well as within the same gland, emphasizing the stochastic nature of thyroid tumorigenesis in this model. The development of larger tumors as seen in female mice is consistent with the well-known sex differences in the occurrence of PTC (Derwahl and Nicula, 2014; Dralle et al., 2015; Rahbari et al., 2010).

To understand the basis of heterogeneous tumor growth, earlier stages of tumor development in non-induced TgCreERT2;BrafCA/+ mice were studied. At 3 months of age, mutant thyroids displayed a limited number of abnormal follicles surrounded by follicles of normal size and shape (Fig. 2A). The enlarged follicles were of two main types: a cell-rich type comprising a thickened and irregularly shaped epithelium, and a hollow type consisting of flattened cells and a distended epithelial lining (Fig. 2A′), hereafter referred to, respectively, as hyperplastic and dilated – including ‘giant’ variant – follicles. On average, each lobe contained two to three clusters of either type of follicular abnormality (Fig. 2B; n=10, serial-sectioned thyroids). Hyperplastic follicles were regularly embedded in the lobe interior, whereas giant follicles had a mostly peripheral location. The neoplastic nature of these early tissue alterations was confirmed by blocking mutant BRAF kinase activity with PLX4720, the precursor compound of PLX4032 or vemurafenib (for V600E mutated BRAF inhibition; we use this acronym synonymously to underline the principal similarity of PLX4720 and PLX4032), administered from the date of weaning, which virtually abolished thyroid enlargement in 3-months-old mice (Fig. 2C). Female and male thyroids showed equal numbers of hyperplastic and dilated follicles, and responded similarly to vemurafenib, although the relative size reduction was more pronounced in females due to the larger glands that develop in untreated mutants (Fig. 2B and C). Notably, vemurafenib inhibited giant follicle formation but did not fully prevent the generation of hyperplastic follicles and microcarcinomas distinguished by a papillary growth pattern (Fig. 2D; see below for further information of the papillary phenotype).

Fig. 2.

Early stages of thyroid tumor development following spontaneous BrafCA activation in TgCreERT2;BrafCA/+ mutant mice. (A) Histogram showing H&E staining of heterotypic follicular abnormalities of the thyroid from a mouse aged 3 months (mo). The boxed area is shown magnified in A′. Asterisks indicate follicles with translucent interior, i.e. lack of colloid. (B-D) Quantitative assessment of changes within the thyroid in response to mutant BRAF kinase inhibition. Dietary pellets with PLX4720 were supplied daily at 417 ppm from 4 weeks onwards and until mice were killed aged 3 months; data were obtained from serial sections. The number of neoplastic follicles in untreated mutants (individual data) is plotted in B. Inhibition of thyroid enlargement (individual data; mean±s.d.; *P<0.005) is plotted in C. Heterogeneous drug response in neoplastic lesions. The mean±s.d. (*P<0.005) of (n): wt (5); untreated mutants (7), drug-treated mutants (9) is plotted in D. (E) Representative image of the thyroid from a mutant mouse aged 6 months, immunostained for thyroglobulin (TG), showing loss of TG in neoplastic lesions. Boxed areas in E are shown magnified in E′ and E′′. Asterisks in E′ and E′′ indicate follicles with altered shape and retained TG in the lumen. L, left lobe; R, right lobe. nf, normal follicle; hf, hyperplastic follicle; gf, giant follicle; arrows indicate hyperplastic epithelium; arrowheads indicate flat epithelium; wt, control wild-type (non-mutant) mice. Scale bars: 500 µm.

Fig. 2.

Early stages of thyroid tumor development following spontaneous BrafCA activation in TgCreERT2;BrafCA/+ mutant mice. (A) Histogram showing H&E staining of heterotypic follicular abnormalities of the thyroid from a mouse aged 3 months (mo). The boxed area is shown magnified in A′. Asterisks indicate follicles with translucent interior, i.e. lack of colloid. (B-D) Quantitative assessment of changes within the thyroid in response to mutant BRAF kinase inhibition. Dietary pellets with PLX4720 were supplied daily at 417 ppm from 4 weeks onwards and until mice were killed aged 3 months; data were obtained from serial sections. The number of neoplastic follicles in untreated mutants (individual data) is plotted in B. Inhibition of thyroid enlargement (individual data; mean±s.d.; *P<0.005) is plotted in C. Heterogeneous drug response in neoplastic lesions. The mean±s.d. (*P<0.005) of (n): wt (5); untreated mutants (7), drug-treated mutants (9) is plotted in D. (E) Representative image of the thyroid from a mutant mouse aged 6 months, immunostained for thyroglobulin (TG), showing loss of TG in neoplastic lesions. Boxed areas in E are shown magnified in E′ and E′′. Asterisks in E′ and E′′ indicate follicles with altered shape and retained TG in the lumen. L, left lobe; R, right lobe. nf, normal follicle; hf, hyperplastic follicle; gf, giant follicle; arrows indicate hyperplastic epithelium; arrowheads indicate flat epithelium; wt, control wild-type (non-mutant) mice. Scale bars: 500 µm.

Mutant thyroids also consisted of normal-sized and moderately enlarged follicles with lumens conspicuously devoid of colloid (Fig. 2A′). A similar effect was evident globally in tamoxifen-induced mice before the compensatory high serum levels of thyroid stimulating hormone (TSH) stimulated goitrous growth (Fig. S3A-C). By contrast, confirming previous notions (Charles et al., 2011), non-induced TgCreERT2;BrafCA/+ mice maintained systemic thyroid hormone homeostasis (Fig. S3C-E). This indicated that the resolution of colloid is likely to be TSH-independent and a direct effect of constitutive mitogen-activated protein kinase (MAPK) signaling in BRAF-mutant thyroid cells.

In mice, thyroglobulin (TG), the thyroid prohormone and main constituent of follicular colloid, is consistently downregulated in BRAFV600E-driven thyroid tumors, featuring a poorly differentiated state (Chakravarty et al., 2011). By using qPCR (Table S2), we confirmed that induced BrafCA activation virtually abolished thyroid-specific gene expression and diminished Pax8, a main transcriptional regulator of thyroid differentiation (Table S2) (Fernández et al., 2015). By contrast, under non-induced conditions, consistent with fewer numbers of oncogene-activated cells, transcription levels of Pax8, Tg, thyroid peroxidase (Tpo) and TSH receptor (Tshr) were partially reduced (Table S2). Morphologically, this corresponded to a markedly heterogeneous tissue consisting of normal follicles with homogeneous TG staining of the colloid, dilated follicles showing retention of agglutinated TG in the lumen and hyperplastic follicles that lack TG immunoreactivity (Fig. 2E and E′). Since microcarcinomas and manifest PTCs were also entirely TG negative (Fig. S4A,B) – confirming loss of differentiation of BRAF-mutant cells – these findings suggested that abnormal follicles with retained TG expression consist of both mutant and normal cells, and do not develop into tumors. Surprisingly, Slc5a5, encoding the sodium-iodide symporter (NIS), was equally suppressed after spontaneous and induced activation of BrafCA (Table S2). It is difficult to explain this other than by NIS expression being more broadly inhibited and encompassing not only BRAF-mutant cells. A bystander effect specifically affecting Slc5a5 might occur, related to the fact that NIS is regulated independently and more sensitive to MAPK activation than other thyroid-specific genes (Chakravarty et al., 2011; Ingeson-Carlsson and Nilsson, 2013).

These findings indicated that mouse thyroid cells respond differently to sporadic activation of BrafCA and give rise to a heterogeneous population of follicular lesions of which only a fraction is tumorigenic. This tissue pattern of neoplastic growth, in fact, reminds of the generation of multinodular goiter that also occurs on the basis of normal follicle heterogeneity related to inherent variations in proliferating capacity of thyroid epithelial cells (Studer et al., 1989).

To get insight about the earliest stages of stochastic tumor development, TgCreERT2;BrafCA/+ mice were crossed with established Cre reporter strains. Consistent with previous findings (Undeutsch et al., 2014), specificity of tracing the thyroid follicular lineage was confirmed by induced activation of Cre in TgCreERT2 mice that carry a lox-STOP-lox lacZ allele targeted to the Rosa26 (R26R) locus (Soriano, 1999). The follicular epithelium of TgCreERT2;R26R mice uniformly stained positive for X-gal corresponding to expression of Cre recombinase, whereas surrounding non-thyroid tissues were unlabeled (Fig. 4A,A′). In the absence of tamoxifen, only occasional X-gal-positive cells appeared in adult mice suggesting a low rate of spontaneous Cre-mediated recombination (Fig. 4B,B′). In similar experiments, the mTmG reporter designed to switch from membrane-tagged mTomato to mGFP expression upon activation was used to increase the resolution of lineage tracing (Muzumdar et al., 2007); hereafter, cells expressing either fluorescent label are referred to as mT+ and mG+ cells, respectively. This revealed a gradual accumulation of mG+ cells from birth eventually comprising almost 40% of the follicular cell population in adult TgCreERT2;mTmG mice (Figs 4C and 5A). The fact that, in aging mice, spontaneous reporter activation leveled off might relate to the accumulation of inactive follicles, comprising cells with diminished function that probably exhibit reduced Tg promoter activity (Studer et al., 1978). Nonetheless, since mTmG faithfully detected contiguous mG+ progenies of normally dividing follicular cells (Fig. 4C), we chose this reporter strain for clonal analysis of BRAF-induced tumor development.

Fig. 4.

Clonal tracing of mutant thyroid cells after spontaneous BrafCA activation. (A,B) X-gal staining of thyroid cells from TgCreERT2;R26R mice injected with tamoxifen (+Tam, A) or not (non-induced, B) to compare activation of the Rosa26 reporter. A′ and B′ are magnified images of labeled cells from the same specimens. Arrowheads in B′ indicate labeled cells. (C) Distribution of normal cells subjected to spontaneous mTmG activation induced by leaky Cre recombinase. (D,E) Expected outcomes when tracing the progeny of BRAF-mutant cells, depending on downregulation of the Cre driver (D) and activation of mTmG before or after that of BrafCA (E). Shown are the sequence of recombination and the corresponding labeling patterns (1-3), of which ‘1.’ represents spontaneous reporter gene activation only. Tg, thyroglobulin transcript; Cre, CreERT2 transcript. (F