Characterization of Thyroid Cancer Driven by Known and Novel ALK Fusions

Abstract

ALK fusions are found in various tumors, including thyroid cancer, and serve as a diagnostic marker and therapeutic target. Spectrum and outcomes of ALK fusions found in thyroid nodules and cancer are not fully characterized. We report a series of 44 ALK-translocated thyroid neoplasms, including 31 identified preoperatively in thyroid fine-needle aspirates (FNA). The average patients’ age was 43 years (range, 8–76 years); only one with radiation history. All 19 resected thyroid nodules with ALK fusion identified preoperatively were malignant. Among nodules with known surgical pathology (n=32), 84% were papillary thyroid carcinomas (PTC) and 16% poorly differentiated thyroid carcinomas (PDTC). PTC showed infiltrative growth with follicular architecture seen exclusively (30%) or in combination with papillary and/or solid growth (37%). Tumor multifocality was seen in 10 (31%) PTC cases. Most PDTC had a well-differentiated PTC component. Lymph node metastases were identified in 10/18 (56%) patients with neck dissection. The most common ALK fusion partners were STRN (n=22) and EML4 (n=17). In 5 cases, novel ALK fusion partners were discovered. All 5 PDTC carried STRN-ALK fusion. On follow-up, 10 patients were free of disease at 2–108 months, whereas two patients with PDTC died of disease. In summary, ALK fusion-positive thyroid carcinomas are typically infiltrative PTC with common follicular growth, which may show tumor dedifferentiation associated with increased mortality. Compared to EML4-ALK, STRN-ALK may be more common in PDTC, and ~10% of ALK fusions occur to rare gene partners. When ALK fusion is detected preoperatively in FNA samples, malignancy should be expected.

Introduction

Thyroid cancer is one of the most quickly increasing in incidence cancer types (Siegel et al. 2019), with papillary thyroid carcinoma (PTC) being the predominant morphologic type. The diagnosis and management of thyroid cancer depends on clinical stage, morphologic type, and molecular profile (Fagin and Wells 2016, Durante et al. 2018). Since the initial discovery and characterization of ALK fusions in thyroid cancer (Kelly et al. 2014, Network 2014, Perot et al. 2014), additional details have emerged. In some prior studies, ALK fusions, most frequently with striatin (STRN) or echinoderm microtubule-associated protein-like 4 (EML4), have been reported in 1–3% of PTC and with a higher frequency in poorly differentiated thyroid carcinomas (PDTC) (Kelly et al. 2014, Network 2014, Chou et al. 2015, Landa et al. 2016, Bastos et al. 2018). Several anaplastic thyroid carcinomas carrying STRN-ALK have also been reported (Kelly et al. 2014, Perot et al. 2014, Godbert et al. 2015). The role of STRN-ALK in the progression from PTC to PDTC thyroid carcinoma has been documented (Nikitski et al. 2018). More recently, new fusion partners of ALK have been identified in thyroid cancer including TFG, GTF2IRD1, and CCDC149 (McFadden et al. 2014, Landa et al. 2016, Vanden Borre et al. 2017). However, the prevalence and spectrum of specific ALK fusion types and impact of ALK fusion partners on tumor phenotype and outcomes remain unknown.

The standard approach to thyroid cancer includes surgery followed by radioactive iodine (RAI) for carcinomas with a higher risk of recurrence. In the management of thyroid nodules and cancers, ALK fusions represent both a diagnostic marker and a potential therapeutic target. The role of ALK as a therapeutic target is being highlighted by a number of case reports of patients with ALK-positive thyroid cancer, including cases of metastatic and RAI-resistant cancer treated with Crizotinib (Demeure et al. 2014, Godbert et al. 2015, Ji et al. 2015, van der Tuin et al. 2019).

We have previously shown that thyroid fine-needle aspirates (FNA) are a reliable source for the detection of ALK fusions, assisting in the diagnosis and management of cytologically indeterminate thyroid nodules (Nikiforova et al. 2018). In prior studies, ALK-translocated thyroid carcinomas were identified by testing resected and histologically diagnosed thyroid carcinomas. The frequency of ALK-translocated thyroid neoplasms prospectively identified by molecular testing of thyroid FNAs is not known and the cytological and histopathological characteristics of these nodules have not been reported.

A significant proportion of ALK-translocated thyroid carcinomas were reported as follicular variant of PTC prior to the reclassification of the encapsulated forms of such tumors as non-invasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP) (Nikiforov et al. 2016). Therefore, whether or not ALK fusions may occur in NIFTP remains unknown.

In this study, we report the detailed analysis of a large series of ALK-translocated thyroid tumors, many of which were identified preoperatively by testing thyroid FNA samples, with particular focus on (i) histopathologic characteristics of tumors, including features of NIFTP, (ii) outcomes when detected preoperatively, and (iii) spectrum of ALK fusion partners and discovery of novel types of ALK fusions.

Methods

Patient Selection and Study Samples

Surgically resected formalin-fixed paraffin-embedded (FFPE) tissue or thyroid FNA samples were used to identify ALK fusions (Tables 1, 2, and Supplemental Table 1). Thirteen patients with ALK-translocated thyroid carcinomas were identified by testing FFPE tissue from surgical resections (Table 1). Surgically resected cases were tested on clinical request, as part of patient’s management. Thirty-one additional patients were identified by testing thyroid FNA samples, 19 of which underwent thyroid surgery (Table 2); 12 of these patients did not have follow-up data available and it is unknown whether thyroid surgery was performed (Supplemental Table 1). Snap-frozen tissue and FFPE tissue were collected at the Department of Pathology, University of Pittsburgh Medical Center (UPMC) following approval by the University of Pittsburgh Institutional Review Board (IRB 991206). This project used the University of Pittsburgh Medical Center, Hillman Cancer Center and Tissue and Research Pathology/Pittsburgh Biospecimen Core shared resource which is supported in part by award P30CA047904. Study was done in accordance with US Federal Policy for the Protection of Human Subjects.

Table 1.

Clinicopathologic summary of patients with ALK fusion identified by testing thyroid resections.

Case #	Age, years	Sex	FNA, BC	Type of surgery	Size, cm	Pathology	pT	pN	Type of ALK fusion	Post- operative RAI	Follow-up, months
1	58	F	n.d.	TT	1.3	iFV	1b	x	e3_STRN-e20_ALK	No	NED, 192
2	61	F	V	TT, CCND	3	mixed	2	0	e3_STRN-e20_ALK	129.1 mCi	NED, 108
3	36	F	IV	TT	1.5	mixed	1b	x	e13_EML4-e20_ALK	75 mCi	NED, 78
4	60	F	V	TT	2	classic	1b	x	e3_STRN-e20_ALK	No	NED, 69
5	72	F	IV	TT	4.2	PDTC	3	x	e3_STRN-e20_ALK	107 mCi	DOD, 37
6	52	M	n.d.	HT	7.5	PDTC	3	x	e3_STRN-e20_ALK	Unknown	DOD, 32
7	29	M	IV	TT, LND	3.3	mixed	2	1b	e3_STRN-e20_ALK	150 mCi	NED, 20
8	33	F	n.d.	TT	6.5	iFV	3	x	e6ab_EML4-e20_ALK	No	NED, 17
9	54	M	VI	TT, LND	3.8	PDTC	3	1b	e3_STRN-e20_ALK	n.a.	n.a.
10	74	F	n.d.	TT	5	PDTC	3	1a	e3_STRN-e20_ALK	n.a.	n.a.
11	16	F	III	TT, CCND	2	mixed	3	1a	e3_STRN-e20_ALK	n.a.	n.a.
12	19	F	n.d.	TT, CCND	1.4	iFV	1b	1a	e3_STRN-e20_ALK	n.a.	n.a.
13	8	M	n.d.	TT, LND	0.9	Solid, classic	1a	1b	e3_STRN-e20_ALK	n.a.	n.a.

F, female; M, male; TT, total thyroidectomy; BC, Bethesda category; n.d., not done; HT, hemithyroidectomy; CCND, central compartment neck dissection; LND, lateral neck dissection; iFV, infiltrative follicular variant; “mixed” refers to a combination of conventional and follicular growth and was ultimately classified as “classic PTC”; PDTC, poorly differentiated thyroid carcinoma; RAI, radioactive iodine therapy; NED, no evidence of disease; DOD, died of disease; n.a., not available

Table 2.

Clinicopathologic summary of patients with ALK fusion identified by testing fine needle aspirates.

Case #	Age, years	Sex	FNA, BC	Type of surgery	Size, cm	Pathology	pT	pN	Type of ALK fusion	Post- operative RAI	Follow-up, months
1	32	F	III	TT, LND	5.4	mixed	3	1b	e3_STRN-e20_ALK	150 mCi	NED, 44
2	43	M	V	TT	1.3	classic	1b	0	e3_STRN-e20_ALK	No	NED, 23
3	32	F	V	TT, CCND	3.7	mixed	2	0	e1_CTSB (UTR)-e17_ALK	No	NED, 2
4	61	F	V	TT	2.2	iFV	2	0	e3_STRN-e20_ALK	53 mCi	NED, 6
5	22	F	III	HT	1.2	mixed	1b	x	e6_EML4-e18_ALK	n.a.	n.a.
6^	76	F	VI	HT	4.2	PDTC	3	x	e3_STRN-e20_ALK	n.a.	n.a.
7	63	F	III	HT	0.5	classic	1a	x	e3_STRN-e20_ALK	n.a.	n.a.
8	34	F	III	HT	3	classic	2	x	e3_STRN-e20_ALK	n.a.	n.a.
9	41	F	III	TT	3	iFV	2	x	e1_CTSB (UTR)-e16_ALK	n.a.	n.a.
10	58	F	III	TT, CCND	3.5	mixed	2	1a	e3_STRN-e20_ALK	n.a.	n.a.
11	12	F	VI	TT, LND	2.2	classic	2	1b	e9_PPP1R21-e20_ALK	n.a.	n.a.
12	69	F	III	TT	0.8	mixed	1a	0	e6_EML4-e18_ALK	n.a.	n.a.
13	28	M	III	HT	3.3	iFV	2	x	e6_EML4-e20_ALK	No	n.a.
14	32	F	IV	TT, CCND	2.8	iFV	2	0	e6_EML4-e20_ALK	n.a.	n.a.
15	47	F	III	HT	2.7	mixed	2	0	e6_EML4-e20_ALK	n.a.	n.a.
16	27	F	III	TT	1	classic	1a	0	e13_EML4-e20_ALK	n.a.	n.a.
17	30	F	III	TT	0.7	classic	1a	x	e20ab_EML4-e20_ALK	n.a.	n.a.
18	24	F	III	HT	1.4	iFV	1b	x	e6_EML4-e20_ALK	n.a.	n.a.
19	41	M	III	TT	2	classic	1b	1a	e13_EML4-e20_ALK	n.a.	n.a.

^{^}This poorly differentiated thyroid carcinoma also harbored TERT C250T, c.−146 C>T mutation

F, female; M, male; BC, Bethesda category; TT, total thyroidectomy; HT, hemithyroidectomy; CCND, central compartment neck dissection; LND, lateral neck dissection; iFV, infiltrative follicular variant; “mixed” refers to a combination of conventional and follicular growth and was ultimately classified as “classic PTC”; PDTC, poorly differentiated thyroid carcinoma; RAI, radioactive iodine therapy; NED, no evidence of disease; n.a., not available

Pathologic Examination

Two authors (S.I.C. and Y.E.N.) reviewed pathology slides. In addition to standard TNM staging parameters (American Joint Committee on Cancer, 8^th edition), growth pattern, histologic variant, presence of infiltrative growth, and criteria for NIFTP were considered (Nikiforov et al. 2016, Nikiforov et al. 2018). Follicular variant was defined as having no appreciable (<1%) papillary component. PDTC were diagnosed following Turin criteria (Volante et al. 2007). For ALK immunohistochemical analysis, 4-μm FFPE sections were hybridized with primary monoclonal antibodies (Anaplastic Lymphoma Kinase, D5F3, Cell Signaling Technology, Inc, Danvers, MA) following the manufacturer’s protocol.

Detection of ALK Fusions

Testing of thyroid FNAs and FFPE-resected thyroid carcinomas as part of clinical care was performed using targeted next-generation sequencing-based ThyroSeq® v3 assay as previously described (Nikiforova et al. 2018, Steward et al. 2018). The assay uses RNA to test for all previously reported types of ALK fusions and also detects the expression levels of ALK tyrosine kinase (TK) domain and of extracellular (EC) domain. Samples with disproportional overexpression of the TK over EC domain of ALK and negative for known types of ALK fusion underwent whole transcriptome analysis (RNA-Seq).

RNA-Seq and Data Analysis

Total RNA was used to prepare the library for RNA exome sequencing with Illumina TruSeqTM RNA Exome Sample Preparation kit v1 (Illumina, Inc, San Diego, CA). The prepared libraries were assessed using an Agilent 2200 TapeStation (Santa Clara, CA), then denatured with sodium hydroxide and loaded on Illumina HiSeq2500. Cluster generation and Paired-end sequencing were performed using HiSeq Paired-End Rapid Cluster kit v2 and HiSeq Rapid SBS kit v2 (Illumina, Inc). The filtered high-quality reads were aligned to the human genome (hg19-GRCh37) using TopHat aligner (Trapnell et al. 2009) and number of reads mapped to each gene was calculated using RSeM (Li and Dewey 2011) and featureCounts (Liao et al. 2014). Gene fusions were detected using Q-score (Li et al. 2008), Chimerascan with custom filtering algorithm (Iyer et al. 2011), Atlas (Huret et al. 2013), CIS (RTCGD) (Akagi et al. 2004), and CHCG (Futreal et al. 2004) databases. The junctions and the individual exon expression levels were used to visualize the fusion point.

RT-PCR and Sanger Sequencing

RNA was subjected to DNase treatment using the Invitrogen DNA-free™ kit (Thermo Fisher Scientific, Waltham, MA). The RNA from frozen tissues was reverse transcribed by Applied Biosystems High Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). The RNA from FFPE tissues and FNA was reverse transcribed by Invitrogen SuperScript IV VILO Master Mix (Thermo Fisher Scientific). RT-PCR was conducted using HotStarTaq DNA Polymerase (Qiagen, Hilden, Germany), and primers listed in Supplemental Table 2. The RT-PCR products were sequenced in both directions using the Applied Biosystems BigDye Terminator Kit and an ABI 3130xl DNA Sequencer (Thermo Fisher Scientific).

Western Blot analysis

Total protein was isolated from frozen tissue homogenized in RIPA buffer (Boston BioProducts, Ashland, MA) using OMNI-GLH (OMNI International, Kennesaw, GA); 40 μg of total protein was resolved by SDS-PAGE and transferred to nitrocellulose membrane (Bio-Rad Laboratories, Hercules, CA). The membranes were incubated overnight at 4˚C with ALK (D5F3, 1:2000, Cell Signaling) or β-actin (1:1000, Cell Signaling) antibodies, followed by anti-rabbit (1:10000, Promega, Madison, WI) HRP-conjugated secondary antibody and signals were detected by ECL (GE Healthcare, Chicago, IL).

Results

Clinical and Morphologic Characteristics

Forty-four patients with ALK-translocated thyroid neoplasms were identified in either surgically removed tumors (n=13) or thyroid FNA samples (n=31). The overall average age of patients was 43.4 years (range, 8 – 76 years), including two pediatric patients, an 8-year-old boy and a 12-year-old girl. Overall, female patients were predominant, 36/44 (82%). The average size of ALK-translocated thyroid nodules was 2.8 cm (range, 0.7 – 7.5 cm). Thyroid FNA was performed on 38 patients and diagnosed in cytology as Bethesda Category (BC) III (25/38, 66%), BC IV (4/38, 10%), BC V (6/38, 16%), and BC VI (3/38, 8%). Additional clinicopathologic and molecular parameters of 44 patients with ALK-translocated thyroid neoplasms are summarized in Tables 1, 2, and Supplemental Table 1. Only one patient had a remote history of therapeutic exposure to RAI, reportedly for multinodular goiter (case #12, Table 2).

Surgical pathology data were available for 32 patients (Tables 1 and 2). The most common procedure was a total thyroidectomy (24/32, 75%). The most common histologic diagnosis was that of well-differentiated PTC (27/32, 84%), followed by 5 cases of PDTC (5/32, 16%)_ENREF_25. The PTC had a predominantly follicular growth pattern with areas of papillary and/or solid growth in 10 cases (10/27, 37%) or without papillary growth meeting the criteria for infiltrative FV PTC (8/27, 30%) (Fig. 1). The remaining 33% (9/27) of PTC had predominantly classic papillary growth pattern with no significant follicular component. Overall, although the proportion of conventional papillary component varied, 19/27 (70%) of PTCs were categorized as classic PTC due to the presence of well-formed papillary structures. Papillae were typically tightly packed and covered by cells with well-developed nuclear features of PTC. The neoplastic follicles were typically irregularly shaped and contained a significant amount of colloid. Nuclear features of PTC were overall less prominent in the areas of follicular growth, with nuclear pseudoinclusions only rarely found. Multifocal PTC was noted in 10 cases (10/32, 31%).

Figure 1. Histopathologic features of representative cases of papillary thyroid carcinoma with ALK fusions.

(A) Papillary thyroid carcinoma with mixed conventional and follicular growth; component of follicular growth was well-circumscribed and thinly encapsulated. H&E, 100x. (B) Interface between follicular (left) and tightly packed papillary (right) components. H&E, 200x. (C) Well-developed nuclear features of papillary thyroid carcinoma (including nuclear pseudo-inclusions, arrow) in the conventional component. H&E, 400x. (D) A minor (<10%) component of solid growth with mitosis (arrow; 1 in 10 high power (200x) fields). H&E, 600x. (E) Lateral neck lymph node with metastatic papillary thyroid carcinoma, mixed follicular and conventional. H&E, 40x. (F) ALK immunohistochemistry, strong diffuse cytoplasmic staining of tumor cells and weak colloid-only background staining in the adjacent normal thyroid. 400x.

PDTC carrying ALK fusions had predominantly solid growth with a minor component of trabecular and insular architecture in some cases (Fig. 2 A, B). Necrosis was identifiable in all cases. The nuclei of PDTC cells were either of smaller size with dark uniformly dispersed chromatin (3 tumors, Fig. 2 C) or larger size with vesicular chromatin (two tumors, Fig. 2 D). Nuclear features of PTC were lost in these areas. Four out of 5 PDTC tumors also contained areas of well-differentiated PTC, either follicular variant or classic type. One PDTC showed an additional genetic alteration, TERT C250T mutation (Case #6, Table 2), whereas all other tumors had an ALK fusion as the only driver mutation identified.

Figure 2. Histopathologic features of poorly differentiated thyroid carcinomas with ALK fusions.

(A) Solid and trabecular growth (H&E, 40x). (B) Areas of tumor necrosis (H&E, 400x). (C) This tumor shows smaller, dark nuclei and significant amount of cytoplasm (H&E, 200x). (D) This tumor has punctate necrosis, larger nuclei with vesicular chromatin, smooth nuclear contours without grooves or pseudoinclusions, and small amount of cytoplasm (H&E, 200x).

Extrathyroidal extension was identified in six cancers (6/32, 19%), four of which were PDTC. There were 11 cases with pT2, eight cases with pT3, and eight cases with pT1b. Lymph node dissection was performed in 18 patients and resulted in pN0 (n=8), pN1a (n=5), and pN1b (n=5). Of 10 patients with lymph node metastases, two patients had PDTC.

ALK IHC was performed on 12 tumors and was positive in 11 cases (11/12, 92%). One case showed indeterminate ALK IHC result and the lack of staining was likely caused by suboptimal tissue preservation in a 10-year old FFPE tissue block.

Clinical follow-up was available for 12 patients (Tables 1 and 2). Five patients received post-operative RAI therapy (dose ranging from 53–150 mCi). Ten patients were free of disease 2 to 108 months after thyroid surgery. Two patients with PDTC developed multiple distant metastases and died of disease 32 and 37 months after surgery.

Preoperative Detection in Thyroid FNA Samples

Thirty-one nodules positive for ALK fusions were identified by testing thyroid FNA samples. This included 15 cases with ALK fusions detected prospectively by ThyroSeq testing of 11,211 consecutive thyroid FNAs with indeterminate cytologic diagnosis over an 11-month period from December of 2017 to September of 2018, yielding the frequency of 0.13%. Overall, among all nodules with ALK fusions first detected in FNA samples, the cytological diagnosis of Bethesda Category (BC) III was established in 23 (74%) cases, BC IV in 1 (3%), BC V in 5 (16%), and BC VI in 2 (7%).

Nineteen nodules positive for ALK fusions in thyroid FNA samples had a known surgical pathology outcome. All 19 were ultimately diagnosed as malignant, including 18 PTC and one PDTC. In all but one case, the carcinoma was diagnosed by primary case pathologists. In Case #5 (Table 2), the 1.2 cm thyroid nodule initially diagnosed as NIFTP was mostly well-circumscribed, but it was lobulated, had complex central fibrous septae outlining the initial primary capsule of a smaller nodule, and a minor conventional component of PTC (Fig. 3 A–C). ALK immunohistochemistry highlighted foci of infiltration at the lesion-normal tissue interface (Fig. 3 D).

Figure 3. Histopathologic features of papillary thyroid carcinoma initially diagnosed as NIFTP.

(case #5, Table 2). (A) This 1.2 cm thyroid mass was mostly well-circumscribed and lobulated. The center of the mass showed complex fibrous septae, perhaps outlining the initial primary capsule of the smaller nodule. H&E, 40x. (B) About 5% of the mass showed branching papillae covered by follicular cells with nuclear features of papillary thyroid carcinoma. H&E, 600x. (C) Periphery of the mass with split thin capsule and permeating tumor cells. H&E, 100x. (D) ALK immunohistochemistry highlighting the focus of infiltration shown in (C), 100x.

Among patients with ALK fusions identified by testing preoperative FNA samples (Table 2), the frequency of known lymph node metastases was 4/19, 21%.

Spectrum of ALK Fusions and Detection of Novel Fusion Partners

The most common fusion partners of ALK were STRN (22/44, 50%) and EML4 (17/44, 39%). In 5/44 (11%) cases, ALK fusions involving novel partners were identified. These cases initially showed a preferential overexpression of the tyrosine kinase domain of ALK over the extracellular domain, suggesting the presence of ALK fusion, with no known fusion types identified. On subsequent RNA-Seq analysis, these cases revealed in-frame ALK fusion to novel partner genes including cathepsin B (CTSB) (n=2), protein phosphatase 1 regulatory subunit 21 (PPP1R21) (n=2), and intersectin 2 (ITSN2) (n=1) (Table 3). Unlike ALK, all of these three genes were found to be expressed in normal thyroid follicular cells, driving the expression of the fused part of ALK coding for the tyrosine kinase domain of the protein.

Table 3.

New types of ALK rearrangements detected in thyroid cancer

Fusion name	Cytogenetic characteristics	Dimerization domain of ALK fusion partner	Encoding ALK exones	Predicted protein size
ITSN2 e29-ALK e18	t(2;2);(p23.3;p23.2)	Coiled-coil	e18-e29	1841 aa
PPP1R21 e9/i9-ALK e20	t(2;2);(p16.3;p23.2)	Coiled-coil	e20-e29	844 aa
PPP1R21 e7-ALK e20	t(2;2);(p16.3;p23.2)	Coiled-coil	e20-e29	794 aa
CTSB e1-ALK e16	t(8;2);(p23.1;p23.2)	None*	e16-e29	711 aa*
CTSB e1-ALK e17	t(8;2);(p23.1;p23.2)	None*	e17-e29	671 aa*

^*Translation start codon in ALK exon16 or ALK exon 17; exon 1 of the partner gene untranslated

Two of these tumors revealed the fusion between exon 1 of CTSB and either exon 17 (Fig. 4) or exon 16 (Supplemental Fig. 1) of ALK. Interestingly, exon 1 of CTSB is part of the 5′UTR and is untranslated, suggesting the presence of an alternative translation site in the chimeric genes. Indeed, alternative translation initiation sites in ALK exons 16 and 17 have been reported (Gasteiger et al. 2003). In the CTSB exon 1-ALK exon 17 fusion, a putative translation start codon is located in ALK exon 17 at chr2:29,450,506–29,450,504 (GRCh37/hg19). The expression of ALK protein in this tumor was confirmed by immunohistochemistry (Fig. 1F) and Western Blot analysis (Fig. 4D). In CTSB exon 1-ALK exon 16 fusion, a putative alternative translation start codon in ALK exon 16 is located at chr2:29,451,837–29,451,835 (GRCh37/hg19).

Figure 4. Identification and confirmation of CTSB (exon 1)-ALK (exon 17) fusion.

(A) Analysis of mRNA expression levels of individual CTSB and ALK exons using RNA-Seq data showing overexpression of ALK exons 17 to 29 downstream of the fusion point. Each box represents an exon and the boxes are colored according to the logarithm of their expression levels as measured in reads per kilobase per million reads. (B) RT-PCR confirmation of CTSB-ALK fusion. L, 100-bp ladder; T, tumor; N, normal tissue; NC, negative control. (C) Sanger sequencing confirmation of CTSB (exon 1)-ALK (exon 17) fusion. Ex, exon; UTR, untranslated region. (D) Western blot analysis of papillary thyroid carcinomas positive (PTC+) and negative (PTC-) for CTSB-ALK fusion and corresponding normal tissue (N). Western blot analysis using ALK antibody showed three bands around 70 to 80 kDa, suggesting the ALK protein translation from more than one start codon in ALK exon 17 or post-translational protein modifications. No ALK protein was detected in normal thyroid tissue or in thyroid tumors lacking this fusion. (E) Schematic representation of the predicted fusion protein retaining the transmembrane domain (TM) and tyrosine kinase (TK) domain of ALK. Propeptide, Peptidase C1A-propeptide; Peptidase C1A, c-terminal; MAM, merpin, A5 protein and receptor protein tyrosine mu; LB, ligand binding domain.

Another two cases showed fusions between PPP1R21 either exon 7 (Supplemental Fig. 2) or exon 9 (Supplemental Fig. 3) and exon 20 of ALK. In the latter case, RT-PCR analysis showed the retention of a 17-nucleotide fragment of PPP1R21 intron 9 at the fusion site, leading to an in-frame mRNA transcript. Finally, a fusion between exon 29 of ITSN2 and exon 18 of ALK was found in one case (Supplemental Fig. 4).

Among 32 cases with surgical pathology data available, there were 18 tumors with STRN-ALK, 11 with EML4-ALK, and 3 with novel types of ALK fusions. All five PDTC cases carried STRN-ALK (p=0.07, Fisher Exact Probability Test). Among PTC, no significant variability in the predominance of follicular, solid, or classic papillary architecture was noted between tumors with different types of ALK fusions.

Discussion

In this study, we provide an in-depth characterization of the largest cohort of patients with ALK-translocated thyroid cancer reported to date. The average age of patients in our series was comparable to that of patients with PTCs unselected for genetic alterations: 43.4 (8–76) versus 46.8 (15–89) years, respectively (Network 2014). The current study describes two pediatric patients, confirming previous observations of this fusion occurring in pediatric patients (Chou et al. 2015, Park et al. 2015, Vanden Borre et al. 2017, Arndt et al. 2018). The overall predilection for female patients appears to be more prominent in ALK-translocated group: 4.5:1 versus 2.7:1 among PTC patients unselected for ALK fusions (Network 2014). The age and sex distribution of ALK-translocated thyroid carcinomas in this study are comparable to those previously reported (Table 4).

Table 4.

Summary of ALK-positive thyroid cancers from prior studies with basic clinicopathologic information

ALK “+” /total, n (%), histologic type	Female/ Male	Age, average (range), years	Morphologic variants and features	Recurrence?	Fusion partner	Reference
1¼98, 2.2%, PTC	11:0	38 (13–68)	Diffuse sclerosing, n=3; iFV, n=6^	No	EML4, n=8	Chou et al 2015
2/129,1.5%, FVPTC	n.a.	n.a.	PTC, n=2	No (n=1)	EML4, TFG	McFadden et al 2014
4/392, 1%, thyroid carcinomas	3:1	37 (13–50)	Classic, n=2; solid, n=1; FV (ETE), n=1;	No	STRN, n=2	Park et al 2015
10/116, 8.6%, PTC	8:2	42 (34–47)	iFV, n=4; classic, n=2; FV PTC, NOS, n=4	Yes, 1/8	EML4, n=6; STRN, n=4	Bastos et al 2018
1/15, PTC	0:1	31	Classic PTC, n=1	No	EML4	Cipriani et al 2017
1 PTC#	0:1	62	TCV PTC	Yes, DM	EML4	Demeure et al 2014
1/12, PTC*	0:1	27	Classic PTC	No	EML4	Pfeifer et al 2019
1/59, PTC^^	0:1	60	Classic PTC	AWD, 5 y	EML4	van der Tuin et al 2019
3/303, 1%, PTC	3:0	10 (7–15)	PTC, pN+, n=1; PTC, NOS, n=2	N.A.	EML4, STRN, GTF2IRD1	Vanden Borre et al 2017
3/75, thyroid carcinomas	1:0§	71	Classic PTC, n=2; ATC, n=1	AWD§	STRN	Perot et al 2014, Godbert et al 2015
3/84,3.5% PDTC; 0/33 ATC	1:2	(30–89)¶	PDTC, n=3	Died, n=2	STRN, EML4, CCDC149	Landa et al 2016
7/77 (9%), PTC~	5:2	23 (16–34)	FV PTC, n=1; PTC NOS, n=6	n.a.	EML4, n=2	Arndt et al 2018

^{^}There were also mixed PTC (n=2), tall cell variant of PTC (n=1), oncocytic PTC (n=1), Warthin-like PTC (n=1). Only EML4, as ALK fusion partner, was specifically tested for.

^#Stable disease after 6 months of Crizotinib treatment

^*Thyroid carcinomas negative for common mutations (e.g., BRAF, RAS)

^{^^}RAI-refractory thyroid carcinomas

^§The sex of two patients was not reported. Response after 6 months of Crizotinib therapy.

^¶Age of individual patients was not reported

^~PTC in radiation exposed patients

ALK-translocated thyroid carcinomas can occur both in patients with and without a history of radiation exposure. However, the prevalence of ALK fusions appears to be higher, up to 7.6–9% (Hamatani et al. 2012, Arndt et al. 2018, Efanov et al. 2018), in PTCs from radiation-exposed patients. An additional approach to increasing the chance of identifying ALK fusions is to exclude thyroid carcinomas with other, more common driver mutations such as BRAF, RAS, and RET/PTC. Of 44 ALK-translocated thyroid carcinomas presented here, only one PDTC showed an additional genetic alteration – TERT promoter mutation. Other studies identified an RAI-refractory conventional PTC with EML4-ALK fusion and TERT mutation (van der Tuin et al. 2019), STRN/ALK-positive PDTC with TERT mutation, and an EML4/ALK-positive PDTC with ATM mutation (Landa et al. 2016). Whole-genome sequencing of an RAI-resistant and recurrent PTC with EML4-ALK fusion revealed co-existing alterations in 55 genes, many of which are likely passenger mutations (Demeure et al. 2014).

Since 2012, about 60 patients with ALK-translocated thyroid cancer have been reported in different studies, ranging from papillary thyroid microcarcinomas to anaplastic thyroid carcinomas, with a significant number of those diagnosed as follicular variant of PTC. Cases of PTC presented in this study and in the literature (Table 4) were evaluated with attention to identifying the most characteristic histology that could help to trigger testing for ALK fusions, and exploring whether the presence of ALK fusion is compatible with the diagnosis of NIFTP.

Morphologically, while not pathognomonic, ALK-translocated PTCs appear to be characterized either by a purely follicular architecture with infiltrative growth, or predominance of follicular growth with islands of tightly packed papillae or classic morphology. Our analysis suggests that tumors with pure classic papillary growth pattern and absence of an appreciable follicular component are less likely to carry ALK fusions. However, we acknowledge that ALK fusions have been reported in tumors with other morphologies, such as diffuse-sclerosing variant of PTC (Table 4). Overall, younger to middle-aged females with infiltrative thyroid cancer showing follicular architecture and negative for BRAF, RAS, and other common genetic alterations are the likeliest to harbor an ALK fusion.

With the growing role of ALK as a diagnostic FNA marker, it was important to establish whether ALK fusion can be seen in benign thyroid nodules or NIFTP (Nikiforov et al. 2016, Nikiforov et al. 2018). To date, 156 follicular adenomas were shown to be negative for ALK fusions (Perot et al. 2014, Park et al. 2015, Bastos et al. 2018, Nikiforova et al. 2018). Here we show that only one of 19 ALK-translocated neoplasms (with ALK fusion first identified in FNA) mimicked NIFTP. This is consistent with the data in the literature. Indeed, out of about 28 PTCs described with sufficient morphologic details (i.e., with comments on variant morphology and invasion), there were no cases of non-invasive follicular variant of PTC (Table 4), rendering the possibility of an ALK-positive NIFTP unlikely.

When follicular growth is predominant and infiltrative growth is subtle, ALK IHC represents a practical tool that helps to asses for invasion more objectively, especially if ALK fusion was already identified by molecular testing of preoperative FNA. In this study, ALK IHC was indeterminate in 1 of 12 cases. In the literature, there are at least two cases with known ALK fusion (identified by FISH or other methods) and negative ALK IHC (Perot et al. 2014, Jeon et al. 2019).

Sixteen percent of ALK-translocated tumors in this study were poorly differentiated, fully meeting the Turin diagnostic criteria for PDTC (Volante et al. 2007). Interestingly, the nuclei of cells in these tumors showed two different types of morphology, correlating with the observation in the animal model of STRN-ALK-driven PDTC (Nikiforov et al. 2019). Most PDTC had a well-differentiated PTC component, suggesting that the tumors initially developed as PTC before undergoing dedifferentiation.

Previously reported cases of ALK-translocated thyroid carcinomas were typically identified using surgically removed tumors. In this study, we explored the frequency of ALK fusions in a large consecutive series of thyroid FNA samples with predominantly indeterminate FNA cytology. We observed that a small but distinct proportion of these aspirates were found to be positive for ALK fusions, and all of those nodules were cancers. Many of these FNAs were diagnosed as Bethesda III by cytology, which is likely due to the presence of abundant colloid and only a moderate degree of expression of nuclear features of PTC in many of these tumors. In another study of ALK-positive tumors, the pre-operative cytology of such nodules was reported in 4 cases and included Bethesda V in 3 and Bethesda III in 1 case (Park et al. 2015).

Two most common fusion partners of ALK were STRN and EML4 in this study and in the literature (Table 4). While we did not identify previously reported ALK fusion partners such as TFG, GTF2IRD1, or CCDC149, we identified fusions with ITSN2, CTSB, and PPP1R21 genes. Of those, ITSN2-ALK and CTSB-ALK fusions have not been previously described. The ITSN2 gene encodes Intersectin 2, a member of multidomain adaptor/scaffold proteins that assemble multimeric complexes implicated in clathrin- and caveolin-mediated endocytosis and rearrangements of the actin cytoskeleton (Predescu et al. 2003, Tsyba et al. 2011). PPP1R21-ALK fusion was previously reported in colorectal carcinoma (Yakirevich et al. 2016) and was not known to occur in thyroid cancer. PPP1R21 is essential for endosome functions (Rehman et al. 2019). In ALK fusions discovered prior to this report, such as STRN-ALK or EML4-ALK, ALK kinase was known to be activated via a ligand-independent dimerization of the chimeric protein mediated by the coiled-coil (CC) domain of the fusion partner with constitutive phosphorylation on Tyr1278 of ALK (Kelly et al. 2014, Holla et al. 2017). Activation through CC domain dimerization is the likely mechanism of ALK activation in PPP1R21-ALK and ITSN2-ALK fusions. In CTSB-ALK, ALK is fused with an untranslated exon 1 of CTSB, with predicted translation starting in ALK exon 16 or 17, suggesting a novel mechanism of ALK expression in these fusions.

The availability of a large number of tumors with STRN-ALK or EML4-ALK fusion allowed us to explore for the first time whether different fusion partners are associated with specific phenotypical or clinical tumor characteristics. A potentially interesting finding was that all 5 PDTC in this study carried STRN-ALK fusions. The number of tumors studied here was still limited and cases of PDTC carrying EML4-ALK have been reported (Landa et al. 2016). Nevertheless, the results of this study raise the possibility that tumors driven by STRN-ALK fusions may have a higher propensity for dedifferentiation. Among PTC, we observed no differences with respect to phenotypical characteristics of tumors driven by these two main types of ALK fusion.

Clinical behavior of thyroid cancer driven by ALK fusions is not well studied. In our series, two patients who died of disease had PDTC with distant metastases, whereas no mortality was observed among patients with PTC on limited follow-up. ALK-driven PTC were multifocal in about one-third of the cases and frequently showed infiltrative growth and lymph node metastases. However, when follow-up was available, 10 patients with PTC were found to be free of disease. In the literature, 23 patients with ALK-translocated thyroid cancer showed no disease recurrence, while distant metastases, persistent disease requiring Crizotinib treatment and deaths of two patients with ALK-translocated PDTC have been reported (Table 4).

In summary, we confirm the occurrence of ALK fusions in a subset of thyroid cancers, specifically in PTC and PDTC, and demonstrate that STRN and EML4 are the two most common ALK fusion partners, with ~10% of ALK fusions are to other rare partner genes. The most typical histologic appearance of ALK-translocated PTC is that of infiltrative follicular variant or predominant follicular growth with a component of papillary growth. We further demonstrated the utility of ALK fusions in identifying thyroid cancer in cytologically indeterminate FNA samples. The presence of an ALK fusion practically rules out the diagnosis of NIFTP. Aggressive cases of ALK-positive thyroid carcinomas are typically PDTC, and the results of this study raise the possibility that tumors driven by STRN-ALK fusions may have a higher predisposition to dedifferentiation (compared to thyroid carcinomas with EML4-ALK fusion). Further studies with higher numbers of distinct ALK fusion partners including the novel ones presented here, and more comprehensive clinical follow-up, are needed to determine the clinical behavior of ALK-driven thyroid carcinomas.