Mouse Model of Thyroid Cancer Progression and Dedifferentiation Driven by STRN-ALK Expression and Loss of p53: Evidence for the Existence of Two Types of Poorly Differentiated Carcinoma

Abstract

Background: Thyroid tumor progression from well-differentiated cancer to poorly differentiated thyroid carcinoma (PDTC) and anaplastic thyroid carcinoma (ATC) involves step-wise dedifferentiation associated with loss of iodine avidity and poor outcomes. ALK fusions, typically STRN-ALK, are found with higher incidence in human PDTC compared with well-differentiated cancer and, as previously shown, can drive the development of murine PDTC. The aim of this study was to evaluate thyroid cancer initiation and progression in mice with concomitant expression of STRN-ALK and inactivation of the tumor suppressor p53 (Trp53) in thyroid follicular cells.

Methods: Transgenic mice with thyroid-specific expression of STRN-ALK and biallelic p53 loss were generated and aged on a regular diet or with methimazole and sodium perchlorate goitrogen treatment. Development and progression of thyroid tumors were monitored by using ultrasound imaging, followed by detailed histological and immunohistochemical evaluation. Gene expression analysis was performed on selected tumor samples by using RNA-Seq and quantitative RT-PCR.

Results: In mice treated with goitrogen, the first thyroid cancers appeared at 6 months of age, reaching 86% penetrance by the age of 12 months, while a similar rate (71%) of tumor occurrence in mice on regular diet was observed by 18 months of age. Histological examination revealed well-differentiated papillary thyroid carcinomas (PTC) (n = 26), PDTC (n = 21), and ATC (n = 8) that frequently coexisted in the same thyroid gland. The tumors were frequently lethal and associated with the development of lung metastasis in 24% of cases. Histological and immunohistochemical characteristics of these cancers recapitulated tumors seen in humans. Detailed analysis of PDTC revealed two tumor types with distinct cell morphology and immunohistochemical characteristics, designated as PDTC type 1 (PDTC1) and type 2 (PDTC2). Gene expression analysis showed that PDTC1 tumors retained higher expression of thyroid differentiation genes including Tg and Slc5a5 (Nis) as compared with PDTC2 tumors.

Conclusions: In this study, we generated a new mouse model of multistep thyroid cancer dedifferentiation with evidence of progression from PTC to PDTC and ATC. Further, PDTC in these mice showed two distinct histologic appearances correlated with levels of expression of thyroid differentiation and iodine metabolism genes, suggesting a possibility of existence of two PDTC types with different functional characteristics and potential implication for therapeutic approaches to these tumors.

Introduction

Thyroid cancer is the most common malignancy of the endocrine system and has shown a steady rise in incidence worldwide in recent decades (1,2). Most thyroid cancers are well-differentiated papillary thyroid carcinomas (PTCs), which overall have a more indolent clinical behavior and five-year survival rates >95%; less common, poorly differentiated thyroid carcinomas (PDTCs) and anaplastic thyroid carcinomas (ATCs) are aggressive cancers with a five-year mortality of 30–50% and >90%, respectively (3–9). Despite the positive effect of surgical treatment and other therapeutic modalities on survival, PDTC and ATC are still frequently lethal malignancies due to their rapid infiltrative growth, propensity for regional and distant spread, and progressive loss of the ability to concentrate radioiodine (RAI) (4,8–14).

PDTC and ATC frequently develop as a result of step-wise dedifferentiation of pre-existing well-differentiated carcinomas (8,10,14–19). PDTC is a particularly interesting group of tumors that have an intermediate position between well-differentiated thyroid cancers and ATC with respect to histopathological features, molecular characteristics, and clinical outcomes (20,21). The diagnostic criteria for PDTC adopted by the WHO classification of endocrine tumors are based on the Turin consensus proposal (20). They include the characteristic solid/trabecular/insular architecture, absence of nuclear features of papillary carcinoma, and presence of a high mitotic rate, convoluted thyroid cell nuclei, or tumor necrosis. Although proposals for calling PDTC based exclusively on high-grade features (tumor necrosis and high mitotic count) exist (22), the Turin-based WHO criteria offer a standardized, reproducible, and clinically relevant approach to diagnosing these tumors (21).

The molecular pathogenesis of PDTC involves the accumulation of genetic alterations that encompass early driver mutations, also shared by well-differentiated thyroid cancers, and late driver mutations, which are characteristic of advanced thyroid cancers including PDTC and ATC (23,24). The most common early driver mutations are point mutations in BRAF and RAS, but gene fusions affecting the ALK, RET, and PPARG genes are also found. Late driver mutations commonly affect TERT and TP53, and less frequently genes encoding the members of the PI3K/AKT signaling pathway, SWI/SNF subunits, and histone methyltransferases (23,24).

PDTC also occupy an intermediate position with respect to expressing the markers of thyroid epithelial differentiation. When studied by immunohistochemistry, PDTC typically preserve immunoreactivity for the transcription factors TTF-1 and PAX8; moreover, they express cytokeratins but can lose the expression of thyroglobulin and epithelial markers such as E-cadherin. Similarly, on gene expression analysis, these tumors have decreased or lost expression of thyroid restricted genes such as SLC5A5 (sodium iodine symporter, NIS) and TG (thyroglobulin), which are essential for thyroid hormone synthesis, and generally preserve the expression of the key thyroid transcription factors PAX8, TTF-1, and FOXE1, as well as cytokeratins (24–30). The expression and functional activity of the NIS protein are directly related to the ability of thyroid cells to trap radioiodine (31–33), which is the most efficient targeted treatment of those thyroid carcinomas that retain iodine avidity (34).

Animal models of thyroid cancer are an important tool for dissecting the molecular mechanisms of cancer progression and exploring new therapeutic approaches in preclinical studies. Several mouse models of tumor dedifferentiation were based on introducing BRAF V600E and other genes affected by point mutations combined with a late event such as p53 inactivation (35–41), and many of them showed rapid progression to ATC, often bypassing the stage of PDTC (42–45). In this respect, thyroid carcinogenesis driven by ALK fusions could represent an attractive model to study PDTC because ALK fusions, most commonly STRN-ALK, are found in different types of human thyroid cancers, including well-differentiated papillary carcinoma, PDTC, and ATC, with the highest incidence (4–9%) in PDTC tumors (24,29,46–48). Further, we have generated a mouse model of thyroid-specific expression of STRN-ALK and showed that this fusion alone can drive the development of PDTC with histopathological features closely recapitulating those of PDTC in humans (49). The aim of this study was to expand this observation and investigate the development of thyroid cancer in mice with thyroid-specific expression of STRN-ALK and inactivation of p53.

Materials and Methods

Generation of STRN-ALK/p53KO mice and experimental design

Experimental triple transgenic Tg-Luc-tdT-STRN-ALK^+/−;Tg-Cre^+/−;Trp53^LoxP/LoxP (STRN-ALK;p53KO) mice with a mixed genetic background (C57BL/6J;FVB/N) were created by crossing already established and characterized Tg-Luc-tdT-STRN-ALK (49), Tg-Cre (50), and Trp53^Loxp (51) mouse lines. Experimental animals were arranged into 6-, 12-, and 18-month-age groups and aged in standard conditions (noG) or treated with goitrogen (G) (2 months on and 1 month off). For goitrogen treatment, mice were supplied with drinking water containing 0.5 g/L methimazole (M8506; Sigma, St. Louis, MO) and 5 g/L sodium perchlorate (410241; Sigma) starting from five weeks of age. For control purposes, the C57BL/6J;FVB/N wild-type and double transgenic Tg-Cre^+/−;Trp53^LoxP/LoxP (p53KO) mice were also generated and aged. Animal care and all experimental procedures were performed in accordance with federal guidelines and institutional policies.

Polymerase chain reaction genotyping and confirmation of thyroid-specific Trp53 ex2-10 deletion

DNA was extracted from tail tissues by using QuickExtract™ DNA Extraction Solution (QE09050; Epicentre, Madison, WI). Genotyping was performed by a multiplex polymerase chain reaction (PCR) using HotStarTaq DNA polymerase (203025; Qiagen, Germantown, MD) and three pairs of primers for simultaneous detection of Tg-Luc-tdT-STRN-ALK, Tg-Cre, Trp53^LoxP, and wild-type Trp53 alleles (Supplementary Table S1). For validation of the thyroid-specific Trp53^loxP ex2-10 deletion, DNA was extracted from snap-frozen thyroid and spleen tissues by using a QIAmp DNA Mini Kit (51306; Qiagen) and PCR was performed as described earlier with primers covering intron 1 and 10 of the Trp53 (Supplementary Table S1).

Ultrasound imaging

To monitor tumor development and growth, ultrasound screening of thyroids in experimental mice was performed monthly by using a preclinical Vevo 3100 micro ultrasound imaging system (FUJIFILM VisualSonics, Toronto, ON, Canada). For high-frequency analysis, 40 MHz B-Mode imaging with the MX550D transducer with an axial resolution of 40 μm was used.

Thyroid tumor samples and serum collection

After euthanasia, blood serum was collected by cardiac puncture and stored at −80°C before measurement of serum thyrotropin (TSH) and free T4 as previously described (52). Guided by the presacrifice ultrasound imaging and tumor-specific red fluorescence from tdTomato, tissue samples from areas of primary tumors with different morphological appearance were dissected and snap-frozen in liquid nitrogen. Remaining tumor tissues were fixed with 10% buffered formalin at 4°C overnight and embedded in paraffin.

Histological and immunohistochemical analysis

Formalin-fixed paraffin-embedded (FFPE) tissues were sectioned at a thickness of 4 μm, deparaffinized, and stained with hematoxylin and eosin (H&E) or immunohistochemically (IHC). For IHC, antigen retrieval and nonspecific blocking were done as previously described (49). Endogenous biotin was blocked with streptavidin (Thermo Fisher Scientific, Waltham, MA) followed by incubation with biotin (Thermo Fisher Scientific) according to the manufacturer's instructions. Primary antibodies (Supplementary Table S2) were applied overnight at 4°C. Secondary biotinylated anti-rabbit antibodies (1:1000, BA-1000; Thermo Fisher Scientific) were applied for one hour at room temperature. After the incubation with complexes of biotinylated peroxidase (Thermo Fisher Scientific) and streptavidin for 20 minutes at room temperature, the visualization of the immunoreactivity was done by using DAB Peroxidase Substrate Kit (Vector Laboratories, Burlingame, CA) according to the manufacturer's instructions.

RNA extraction, quantitative RT-PCR, RNA sequencing, and gene expression analysis

Tumor samples with verified distinct histology were subjected to RNA extraction. Total RNA was extracted by using TRIzol reagent (Invitrogen, Carlsbad, CA) from frozen thyroid tissues, followed by treatment with DNAse I (Invitrogen) and purification with RNeasy Mini Kit (Qiagen). RNA from FFPE tumor samples was extracted by using an RNeasy FFPE kit (Qiagen) after H&E-guided microdissection of tumor regions of interest.

For quantitative RT-PCR (qRT-PCR), the purified RNA was reverse transcribed by using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA). Samples were tested in triplicate by using a 7500 Real-Time PCR System (Applied Biosystems) with QuantiTect SYBR^® Green PCR Kit (Qiagen) and primers listed in Supplementary Table S1. Expression levels of each gene relative to β-actin (Actb) were calculated by using Q-Gene software (53). To minimize biases associated with goitrogen treatment, all gene expression data from tumor samples were normalized to the average expression levels of corresponding genes in nontreated (n = 5) or goitrogen-treated (n = 5) benign thyroid samples.

For RNA sequencing, RNA extracted from frozen tissues was subjected to library preparation by using NEBNext^® Ultra™ II RNA Library Prep Kit for Illumina (NEB, Ipswich, MA) and paired-end sequenced on the Illumina NovaSeq 6000 S4 flow cell (20M PE150 reads) by Novogene (Novogene Corporation, Sacramento, CA). Raw sequencing data were imported into CLC Biomedical Genomics Workbench 12 (Qiagen), high-quality trimmed paired reads were aligned to the mouse genome (Ensembl v80), and the number of reads mapped to each gene was normalized by using transcripts per million. The principal component analysis (PCA) was performed on whole-transcriptome data. For differential gene expression (DEG) analysis, gene expression data from PDTC type 1 (PDTC1) (n = 3) and PDTC type 2 (PDTC2) (n = 3) tumor samples were compared with gene expression data in benign (noG = 2, G = 2) thyroid samples while controlling for goitrogen treatment. Genes with an absolute fold change >3 and false discovery rate (FDR) p-value of <0.05 were selected as differentially expressed genes and used for building the Venn diagram. A heatmap of the expression of 15 preselected thyroid differentiation genes was generated by using complete-linkage clustering (Manhattan distance). ERK-score calculation was performed as previously described (47). Gene expression datasets were additionally analyzed by using GSEA (MSidDB database, http://software.broadinstitute.org/gsea/index.jsp) and Ingenuity Pathway Analysis (Qiagen) software.

Statistical analysis

Statistical analysis was performed by using IBM SPSS Statistics version 21 (IBM, Armonk, NY). p-Values were two sided and considered significant if ≤0.05. DEG analysis was performed by using multifactorial statistics based on a negative binominal generalized linear model (by default in CLC Biomedical Genomics Workbench 12).

Results

Generation of mice with thyroid-specific expression of STRN-ALK and loss of p53

To investigate the effect of STRN-ALK expression combined with a common late event, inactivation of p53 function, we created triple transgenic mice by crossing the existing Tg-Luc-tdT-STRN-ALK, Tg-Cre, and Trp53^Loxp mouse lines (Fig. 1A, B). Final Tg-Luc-tdT-STRN-ALK^+/−;Tg-Cre^+/−;Trp53^LoxP/LoxP (STRN-ALK;p53KO) experimental animals had thyroid-specific expression of the HA-tagged STRN-ALK oncoprotein driven by the transgenic thyroglobulin promoter (Supplementary Fig. S1) and Cre-recombinase-mediated thyroid-specific deletion of LoxP-flanked exons 2–10 of Trp53^LoxP alleles (Fig. 1C), introduced in a homozygous state.

FIG. 1.

STRN-ALK;p53KO mice generation and validation of the genotypes. (A) Simplified overview of mouse breeding and scheme of transgenic alleles and their interaction resulting in the STRN-ALK;p53KO phenotype. tpTg, transgenic thyroglobulin promoter; Cre encodes transgenic Cre recombinase; Luc-tdT encodes transgenic fusion protein of Luciferase (not measured in this study) and tdTomato (red fluorescence was used for detection of lung metastases during dissection); T2A, “self-cleaving” peptide sequence; HA-STRN-ALK encodes transgenic HA-tagged STRN-ALK; epTrp53, endogenous Trp53 promoter; Trp53 encodes endogenous TRP53 (p53). (B) Representative results of multiplex PCR used for verification of the genotypes during the generation of experimental mice. (C) PCR confirmation of thyroid-specific deletion of exons 2–10 in Trp53^loxP alleles in STRN-ALK;p53KO mice. PCR, polymerase chain reaction. Color images are available online.

STRN-ALK;p53KO mice develop metastatic and lethal thyroid tumors with high penetrance

STRN-ALK;p53KO mice were divided into 6-, 12-, and 18-month-age groups (n = 11–16 per age/treatment group) and housed on normal diet or on goitrogen treatment, which led to a ∼3-fold decrease in FT4 and a ∼100-fold rise in TSH levels (Supplementary Fig. S2). Thyroid tumor development and progression in these mice were monitored by using ultrasonographic screening. Ultrasound imaging allowed early detection of ∼1-mm thyroid nodules from the age of seven and five months in mice on regular diet (noG) or treated with goitrogen (G), respectively. On initial detection, continuous monitoring of tumor nodules showed their gradual increase in volume and, in many cases, transition to a phase of rapid growth accompanied by changes in ultrasonographic patterns. Subsequent histological examination of such cases revealed areas of PDTC growing adjacent to PTC (Fig. 2A, B). Visual and manual examination of the mice generally revealed large hard cervical masses fixed to the surrounding tissues.

FIG. 2.

Ultrasonographic monitoring, survival, and tumor phenotype of STRN-ALK;p53KO mice. (A) Ultrasonography-based follow-up of tumor development and progression. (A, B) Correlation of ultrasonographic patterns with histopathological findings. The area of PDTC shows marked hypoechogenicity compared with PTC. (B) Representative microscopic H&E images showing the co-occurrence of PTC and PDTC in the same thyroid lobe. (C) Representative microscopic H&E images of ETE into the larynx, TI and VI. (D) Representative gross-necropsy image showing cervical tumor mass, dislocation of the trachea (black arrow), and red fluorescence (from tdTomato encoded by same transgenic allele as HA-STRN-ALK) of the primary tumor and lung metastases (white arrows). (E) Representative image of PDTC lung metastases positive for TG and HA-tagged STRN-ALK (HA). ETE, extrathyroidal extension; H&E, hematoxylin and eosin; PDTC, poorly differentiated thyroid carcinoma; PTC, papillary thyroid carcinoma; TG, thyroglobulin; TI, tracheal; VI, vascular invasion. Color images are available online.

Thyroid cancers were first detected on histological analysis of six-month-old mice on goitrogen treatment (Table 1). With continued aging, about half of tumor-bearing animals (noG, n = 7; G, n = 10) were removed from the experiment before the planned sacrifice date due to the development of life-threatening conditions. By the age of 12 months, 48% of mice on regular diet and 86% mice treated with goitrogen developed histologically confirmed thyroid cancers. In mice treated with goitrogen, the distribution of histological cancer types was shifted to dedifferentiated tumors (Table 1). Among age-matched controls, wild-type (noG, n = 17; G, n = 11) and p53KO (noG, n = 22; G, n = 10) mice developed no cancer.

Table 1.

Cancer Frequencies in STRN-ALK;p53KO Mice

			Cancer typeb
Goitrogen	Age group,amonths	Total number of mice	PTC, n (%)	PDTC, n (%)	ATC, n (%)	Any type, n (%)	Lung metastases, % of primary tumors
−	6	13	—	—	—	—	—
	12	23	11 (48)	5 (22)	3 (13)	11 (48)	1 (9%)
	18	7	5 (71)	2 (29)	1 (14)	5 (71)	2 (40%)
+	6	20	3 (15)	3 (15)	—	5 (25)	1 (20%)
	12	14	7 (50)	11 (79)	4 (29)	12 (86)	4 (33%)
	18	1	—	—	—	—	—

^aDue to the early removal events, age groups are based on time of sacrifice.

^bSome mice had more than one cancer type.

ATC, anaplastic thyroid carcinoma; PDTC, poorly differentiated thyroid carcinoma; PTC, papillary thyroid carcinomas.

Irrespective of goitrogen treatment, thyroid cancers in these mice showed histological features of aggressive behavior, including extrathyroidal growth and invasion into trachea and blood vessels (Fig. 2C). Lung metastases were found in 8 (24%) tumor-bearing mice and were presented as multiple small nodules (Fig. 2D, E and Table 1). Most metastatic tumors either had PDTC histology or contained areas with features of PTC and PDTC.

Thyroid tumors in STRN-ALK;p53KO mice undergo step-wise dedifferentiation

Thyroid tumors developed in these mice included well-differentiated PTC (n = 26), PDTC (n = 21), and ATC (n = 8), which frequently coexisted in the same thyroid gland, although they typically grew as well-defined tumor nodules (Table 2 and Fig. 3). The co-occurrence of PTC and PDTC was the most common (n = 11), but all other combinations of PTC, PDTC, and ATC were also observed (Table 2).

Table 2.

Thyroid Cancer Types in All Age Groups of STRN-ALK;p53KO Mice

	Cancer type
Goitrogen	PTC only	PDTC only	PTC and PDTC	PTC and ATC	PDTC and ATC	PTC and PDTC and ATC
−	8	0	4	1	0	3
+	2	4	7	1	3	0
Total	10	4	11	2	3	3

FIG. 3.

Histological and immunohistochemical features of different stages of thyroid cancer dedifferentiation seen in STRN-ALK;p53KO mice. ATC, anaplastic thyroid carcinoma; fvPTC, follicular variant of papillary thyroid carcinoma; hvPTC, hobnail variant of papillary thyroid carcinoma. Color images are available online.

PTCs had microscopic features of different tumor variants, including follicular variant, classic type of papillary carcinoma, and hobnail variant. Although PTC was overall the most common tumor type found, in many mice it was seen as a smaller component; whereas PDTC tumors were typically most dominant, occupying the majority of the thyroid volume and reaching a size of 8–15 mm (Fig. 3). Histologically, all PDTC tumors had microscopic characteristics similar to those seen in PDTC in humans as defined by the Turin criteria. They commonly showed a solid growth pattern, although insular and trabecular architecture was also observed. Nuclear features of papillary carcinoma were lost, and high mitotic activity and tumor necrosis were frequently seen. Eight cases showed anaplastic transformation; however, the anaplastic component was not prominent and was present as small distinct foci (maximum size of 2–3 mm) of ATC adjacent to PDTC (six out of eight ATC) or next to PTC (two out of eight ATC). The ATC tumors typically had a pleomorphic cell appearance, but spindle cell and squamoid patterns were also seen.

Immunohistochemical analysis showed that PTC and PDTC tumors had strong expression of HA-tagged STRN-ALK whereas cells in ATC generally had focal weak or no expression of the transgene (Fig. 3). All PTCs and PDTCs preserved the expression of the transcription factors TTF-1 and PAX8, with typically higher levels in PTCs. All PTCs and approximately half of PDTCs preserved immunoreactivity for thyroglobulin. ATC lost expression of all of these thyroid differentiation markers. Expression of epithelial markers cytokeratins 17/19 and E-cadherin showed a gradual decrease and complete loss with tumor dedifferentiation, with the opposite trend for vimentin, a mesenchymal marker.

PDTC tumors in STRN-ALK;p53KO mice show two subtypes with distinct histological and immunohistochemical features

Further histological analysis of the PDTC tumors led us to identify two distinct microscopic tumor appearances that correlated with the expression of thyroid differentiation markers detectable by immunohistochemistry (Fig. 4). Although all of these PDTC tumors fully met the Turin histological criteria of PDTC, 12 tumor nodules were composed of solid sheets of cells that retained a significant amount of cytoplasm and had smaller-sized nuclei with dense, evenly distributed chromatin; we defined these tumors as PDTC1 (Fig. 4D). We also found 13 PDTC tumor nodules composed of solid sheets of cells with a reduced amount of cytoplasm and larger, vesicular nuclei that had prominent and frequently multiple nucleoli; these tumors were designated as PDTC2 (Fig. 4D). In the majority of mice, the PDTC tumors had histological features of either type 1 (n = 8) or type 2 (n = 9); whereas in four cases, the mouse thyroid glands had nodules with PDTC1 and PDTC2 appearance coexisting as distinct, well-delineated nodules located next to each other (Fig. 4B, C). Foci of anaplastic transformation were found adjacent to PDTC2 tumors in five cases and adjacent to a PDTC1 tumor in one case. The distribution of PDTC1 and PDTC2 tumors was similar in mice with and without goitrogen treatment.

FIG. 4.

Ultrasonographic, microscopic, and immunohistochemical features of PDTC1 and PDTC2 in STRN-ALK;p53KO mice. (A) Ultrasonographic image of the thyroid gland with PDTC1 and (B) subsequent appearance of PDTC2 nodule within one month. (B, C) Correlation between ultrasonographic and histological appearances of PDTC1 and PDTC2. Type 2 PDTC with more solid morphology shows lower echogenicity compared with type 1 PDTC. (D) Representative microscopic H&E images showing histological differences between type 1 and type 2 of PDTCs. High-power inserts represent differences in nuclear-to-cytoplasm ratio and nuclear features between two types of PDTC. (E) Representative microscopic H&E and immunohistochemistry images showing preserved expression of HA-tagged STRN-ALK and thyroid differentiation marker TTF-1 in both types of PDTC; and loss of thyroglobulin and E-cadherin immunoreactivities in PDTC2. PDTC1, poorly differentiated thyroid carcinoma type 1; PDTC2, poorly differentiated thyroid carcinoma type 2. Color images are available online.

Although immunohistochemical analysis revealed that both PDTC types had comparable levels of STRN-ALK transgene expression, evaluated by staining with the anti-HA tag antibody, the immunoreactivity for several thyroid cell markers showed significant differences between these PDTC types (Fig. 4E). Specifically, PDTC1 tumors consistently retained typically diffuse and moderately strong or weak staining for thyroglobulin and E-cadherin, whereas both stainings were frequently lost in PDTC2 tumors. TTF-1 and PAX8 immunoreactivity was preserved in both PDTC types, although in some PDTC2 tumors it was weaker.

A review of ultrasound images corresponding to histologically diagnosed PDTC1 and PDTC2 nodules obtained during different time points allowed us to trace back the appearance and growth of these tumors. In those cases where both tumor types were present in one thyroid gland, the analysis revealed that PDTC1 tumors appeared at an earlier age as compared with PDTC2 (Fig. 4A–C).

Type 1 and type 2 PDTC tumors have different expression levels of thyroid differentiation genes

To further explore potential differences between PDTC tumors with type 1 and type 2 morphology, we investigated gene expression profiles of these tumors by using RNA-Seq. We generated gene expression data for six PDTC tumors (PDTC1, n = 3; PDTC2, n = 3) as well as two well-differentiated PTC and four benign thyroids (noG, n = 2; G, n = 2). The unsupervised PCA of tumor transcriptional profiles showed that PDTC and PTC were clustered separately from each other and from benign thyroids. Moreover, type 1 and type 2 PDTC tumors also appeared to cluster separately (Fig. 5A).

FIG. 5.

Distinct gene expression profiles of type 1 and type 2 PDTC tumors. (A) Principal component analysis of whole-transcriptome data obtained from type 1 PDTC (n = 3), type 2 PDTC (n = 3), and PTC (n = 2) tumor samples and benign (regular diet [BnoG], n = 2; goitrogen [BG], n = 2) thyroids. Each dot corresponds to a particular sample. (B) Venn diagram showing differentially expressed genes (absolute fold change >3, FDR p-value of ≤0.05) in PDTC1 (n = 3) and PDTC2 (n = 3) compared with benign thyroids (n = 4). (C) Cluster analysis and heatmap of 15 thyroid differentiation genes expression in PDTC1 (n = 3) and PDTC2 (n = 3) tumor samples. (D) Box-plots showing relative expression of main thyroid differentiation genes in PDTC1 (n = 5) and PDTC2 (n = 5) normalized to benign thyroids (regular diet, n = 5; goitrogen, n = 5) while controlling for goitrogen treatment. Data presented as % of corresponding gene expression in benign thyroid tissues. Boxes represent 25th–75th percentiles; Whiskers represent maximum and minimum values. *p < 0.05, **p < 0.005, Mann–Whitney test. FDR, false discovery rate. Color images are available online.

Gene expression analysis in PDTC1 and PDTC2 tumors was performed in comparison to benign thyroid tissues (absolute fold change >3; FDR p ≤ 0.05) (Supplementary Table S3A). The analysis revealed significant differences in gene expression profiles between type 1 and type 2 PDTC tumors (Fig. 5B). Further analysis of gene expression sets from these tumors showed that PDTC2 tumors had a significant upregulation of pathways related to cell cycle control and mitosis, downregulation of genes associated with the epithelial phenotype (Supplementary Fig. S3A, B), and enrichment in other pathways including thyroid hormone biosynthesis (Supplementary Table S4A–C).

Next, we focused on the analysis of genes involved in thyroid differentiation, iodine uptake, and metabolism. As compared with benign thyroid tissues, the expression levels of several thyroid genes involved in iodide metabolism and thyroid hormone synthesis such as Tg, Tpo, Ano1, and Duox2 were significantly decreased in PDTC2 but not in PDTC1 (Supplementary Table S3B). Both types of PDTC had significantly lower expression of Slc5a5 (Nis) as compared with benign thyroid tissue, but the average expression was approximately twofold lower in PDTC2. Some of the other thyroid differentiation genes (Tshr, Foxe1, and Glis3) showed a tendency for a lower expression in PDTC2; in clustering analysis of thyroid differentiation genes, PDTC1 and PDTC2 tumors formed two distinct clusters (Fig. 5C). To test whether the observed downregulation of thyroid differentiation gene expression is associated with an increased MAPK pathway activity, we calculated the ERK-score, which showed a tendency to be higher in PDTC2 tumors, although the number of studied samples was low (Supplementary Fig. S3C).

To validate the RNA-Seq data, we performed qRT-PCR analysis to compare the expression of key thyroid follicular cell differentiation and iodide metabolism genes between five PDTC1 and five PDTC2 tumors. The analysis showed that PDTC2 tumors had significantly lower expression of the Slc5a5 (Nis), Tg, Tpo, and Duox2 genes; whereas the expression of Tshr, Nkx2-1 (Ttf-1), Pax8, and Foxe1 in these tumors displayed a tendency to decrease, which, however, did not reach statistical significance (Fig. 5D). These findings provide further evidence supporting the existence of two types of PDTC in STRN-ALK;p53KO mice, which have not only distinct histological and immunohistochemical characteristics but also different transcriptomic profiles, including genes that are of relevance for a differentiated thyroid phenotype.

Discussion

In this study, we generated a new mouse model of thyroid-specific expression of STRN-ALK and inactivation of p53 and showed that these animals develop metastatic and frequently lethal thyroid cancers with a step-wise progression from PTC to PDTC and ATC and phenotypical characteristics recapitulating thyroid cancers in humans. Further, we demonstrate that in this mouse model the PDTC tumors have two distinct morphological appearances that correspond to different levels of expression of thyroid differentiation genes.

Fusions of the ALK oncogene are found in various types of thyroid cancers, with a higher incidence in dedifferentiated tumors and particularly in PDTC as compared with well-differentiated thyroid PTC (24,46–48). In a recent study, we have confirmed the oncogenic properties of the most common type of ALK fusion, STRN-ALK, and demonstrated that mice with thyroid-specific expression of STRN-ALK develop PDTC (49). Based on the fact that human PDTC and ATC frequently harbor additional late mutations such as alterations of TP53 (23,24), we expanded our mouse model of STRN-ALK-driven PDTC by adding a thyroid-specific biallelic inactivation of p53. Loss of p53 function in these mice leads to a higher penetrance with respect to generation of PDTC and progression to complete tumor dedifferentiation with formation of ATC.

The observed patterns of co-occurrence of tumors with different degrees of differentiation and a shift to more dedifferentiated tumors in older animals, especially after treatment with goitrogen, recapitulate the accepted paradigm of step-wise dedifferentiation of thyroid cancer in humans. In includes the progression of well-differentiated cancer such as PTC to less differentiated PDTC and fully dedifferentiated ATC, a process that is accompanied by progressive accumulation of driver mutations and a loss of tumor suppressors (54,55). This progression model is supported by the co-occurrence of well-differentiated and/or poorly differentiated cancer areas in 30–80% of ATC in humans (12,16,56–61). Additional and more direct evidence is offered by animal models of thyroid cancer driven by an early event such as activating BRAF or RAS mutations or inactivation of Pten combined with the loss of p53 or Nf2 tumor suppressor genes or activating mutation in PIK3CA. These animals either develop thyroid cancer with rapid anaplastic transformation (42–45) or show cancer dedifferentiation with a well-defined step of PDTC (35–41).

The thyroid tumors present in this mouse model of thyroid-specific STRN-ALK expression coupled with p53 inactivation are morphologically and functionally similar to the respective cancer types seen in humans. Interestingly, among PTC tumors, we observed classical papillary type, follicular variant, as well as hobnail variant tumors. The latter resembled the hobnail variant of PTC in humans that frequently carry mutations in TP53 and show progression to PDTC (62–66). In many of these mice, PDTC was the dominant tumor type found on histological examination. These tumors had all morphologic characteristics defined for human PDTC by the Turin criteria that had been adopted by the WHO classification of endocrine tumors (20,21). Similar to human tumors, PDTC in these mice show a variable decrease or loss of immunoreactivity for thyroglobulin and epithelial markers such as E-cadherin and preserved PAX8 and TTF-1 thyroid differentiation markers, consistent with an intermediate immunophenotype between well-differentiated cancer and ATC observed in humans (25,26,28,29,67,68).

However, despite fully meeting the Turin diagnostic criteria, PDTC tumors in STRN-ALK;p53KO mice revealed two distinct tumor cell appearances with respect to the amount of cytoplasm, size of nuclei and appearance of chromatin, and nucleus to cytoplasm (N:C) ratio. Tumors designated as PDTC1 are composed of cells with smaller hyperchromatic nuclei, a larger amount of cytoplasm, and an overall lower N:C ratio. This cellular appearance is more reminiscent of well-differentiated thyroid cancer cells, and indeed, these tumors preserve stronger immunoreactivity for thyroglobulin and E-cadherin and higher levels of expression of mRNA of genes related to thyroid differentiation and iodine metabolism. In contrast, PDTC2 tumors are composed of cells with larger nuclei, vesicular chromatin with prominent and typically multiple nuclei, a smaller volume of cytoplasm, and a higher N:C ratio. These cells have virtually no resemblance to differentiated thyroid cancer cells, but they retain a monomorphic appearance of the nuclei without the marked atypia seen in anaplastic carcinoma cells. Compared with PDTC1, PDTC2 tumors show a more pronounced loss of markers of thyroid differentiation on immunohistochemistry and gene expression analysis. However, both types of PDTC found in this study do not differ with respect to predominantly solid growth pattern, high mitotic activity, and presence of coagulative necrosis, which are diagnostic features of PDTC in humans.

Most tumor-bearing thyroid glands contained either PDTC1 or PDTC2 tumors, which frequently were located adjacent to areas of well-differentiated PTC. However, in four cases, both PDTC types were found coexisting in the same thyroid. Reconstruction of tumor nodule growth using ultrasound images taken at different time intervals suggests the possibility of progression from PDTC1 to PDTC2. Further, in most cases, areas of ATC were found adjacent to type 2 PDTC. These findings point toward the possibility of progression from PDTC1 to PDTC2 and then to ATC. Nevertheless, in two cases, the areas of ATC were seen adjacent to PTC without appreciable PDTC tumor present, which suggests that a direct transition from well-differentiated PTC to ATC may also occur. Overall, based on the frequencies of various tumor types coexisting in the same thyroid, it appears that dedifferentiation of thyroid cancer driven by activated ALK kinase coupled with loss of p53 may often involve a PTC → PDTC1 → PDTC2 → ATC progression, although deviations from such linearity or more direct dedifferentiation pathways are also possible (Fig. 6).

FIG. 6.

Putative scheme of progression and dedifferentiation of thyroid cancer in STRN-ALK;p53KO mice. Thickness of arrows represents number of cases. Color images are available online.

Differences in expression of Nis and other genes involved in thyroidal iodide uptake and organification found between the two PDTC types in these mice are intriguing and potentially relevant. In humans, PDTCs are known to have significant variability in iodine uptake and response to RAI. In fact, most studies showed that approximately half of PDTCs in humans retain substantial and occasionally a strong ability to concentrate RAI, whereas the other half are RAI-refractory (4,69,70). The results of this study raise at least a theoretical possibility that the group of PDTC defined by the current diagnostic criteria is not homogeneous and, although still occupying an intermediate position in terms of thyroid differentiation between well-differentiated cancer and ATC, may include two tumor types with different degrees of expression of iodide-metabolizing genes and responses to RAI. However, it remains unclear whether the same PDTC types can be found in humans and whether or not the differences in expression of thyroid differentiation genes would translate into variable abilities of these tumor cells to concentrate and retain radioiodine.

We observed that goitrogen treatment accelerated thyroid tumor development and progression from PTC to PDTC and ATC. This is likely caused by the elevated TSH levels and, surprisingly, preserved expression of Tshr in the majority of mouse PDTC tumors. This observation is consistent with other studies showing that elevated levels of TSH play a role in the initiation and progression of thyroid cancers and are associated with a more severe tumor phenotype in mice (71,72) and humans (73–75).

This mouse model has a limitation related to the expression of STRN-ALK controlled by a transgenic thyroglobulin promoter, which results in a significant decrease or disappearance of transgene expression during anaplastic transformation. This may be responsible for a lower penetrance with respect to the formation of ATC, and it make these mice unsuitable to study the effect of targeted ALK kinase inhibition. The generation of a new doxycycline-inducible STRN-ALK mouse line is ongoing, which should overcome the differentiation-dependent limitation of the current model. In addition, the number of tumor samples used for histological and molecular characterization of the two groups of PDTC tumors was limited and the ability of these tumors to concentrate and retain RAI was not analyzed in this study. Therefore, it would be important to expand the analysis to a larger cohort of PDTC samples and evaluate RAI uptake by PDTC1 and PDTC2 tumors in vivo or by using established tumor cell primary cultures.

In summary, the results of our study demonstrate that expression of STRN-ALK with simultaneous loss of p53 function in murine thyroid cells initiates a program of thyroid carcinogenesis and multistep dedifferentiation with evidence of progression from PTC to PDTC and ATC. Further, our findings raise the possibility of the existence of two PDTC types with different levels of thyroid differentiation and expression of iodine-metabolizing genes. These findings may prove to be relevant to human PDTC, which requires further investigation.

Author Disclosure Statement

No competing financial interests exist.

Supplementary Material

Supplementary Table S1

Supplementary Table S2

Supplementary Table S3

Supplementary Table S4

Supplementary Figure S1

Supplementary Figure S2

Supplementary Figure S3