Introduction

The Eukaryotic Translation Initiation Factor 1A, X-Linked (EIF1AX) gene codes for the eukaryotic translation initiation factor 1A (eIF1A), an essential component of the 43S pre-initiation complex (PIC). eIF1A stabilizes the binding of the ternary complex (Met-tRNAi:eIF2:GTP) to the 40S ribosomal subunit to form the 43S PIC which is then recruited to the 5’ end of capped mRNA to form the 48S pre-initiation complex. This complex subsequently starts the scanning process to locate the start codon and initiate protein translation [1, 2]. EIF1AX mutations alter the RNA-binding surface of eIF1A and result in defects in 43S and 48S preinitiation complex formation which interfere in protein translation [3]. Deregulation of translation initiation is common in tumorigenesis [4]. Furthermore, overexpression of E1F1AX increases the expression of Cyclin D1, a cell cycle regulator, triggering cell proliferation in vitro [5]. Mutations in EIF1AX gene were initially discovered in uveal melanomas; they have since been reported in other cancers including low-grade gliomas, lung, uterine, ovarian and papillary thyroid carcinomas (PTC) [6,7,8,9]. In thyroid, mutations in the EIF1AX gene were deemed to be driver events for thyroid carcinogenesis by the Thyroid Cancer Genome Atlas study [9], which reported EIF1AX mutations in 1.5% of well differentiated thyroid carcinomas with near-mutual exclusivity with other genetic abnormalities. Subsequent studies have confirmed the presence of EIF1AX mutation in around 1–2% of PTCs [9, 10], 5% of FTCS [11] and 10% of PDTCs and ATCs [12, 13]. The TCGA study was however limited by population selection and did not include benign lesions [9]. This mutation has since been confirmed in benign thyroid nodules [10, 14,15,16].

Significant variations in the risk of malignancy (ROM) conferred by isolated EIF1AX mutation are observed across studies, ranging from 13% to approximately 50% [10, 17,18,19] but tumors are typically low-risk [9, 10]. The ROM is higher in EIF1AX splice site mutation and when EIF1AX mutation co-exists with other molecular alterations such as TP53 or RAS mutation where it correlates with aggressive phenotypes [4, 9, 10, 12,13,14, 20]. In light of the new World Health Organization fifth edition classification of endocrine tumors (WHO 5th ed.) [21], we explored the occurrence of EIF1AX mutation in thyroid lesions in two settings; we first report our experience with thyroid nodules with an indeterminate cytology and EIF1AX mutation as encountered in the general population at a tertiary hospital (YNHH). As the association of EIF1AX with malignant tumors has continually been underscored [9,10,11,12,13,14,15], we also sought to examine the distribution of EIF1AX mutation in various thyroid carcinomas by expanding our study to a separate set of patients with thyroid tumors from a referral cancer center (MSKCC) to see if the genetic abnormalities and their phenotypic correlates are influenced by institutional practices.

Materials and methods

Study patients

This study was approved by the Yale University and Memorial Sloan Kettering Cancer Center (MSKCC) Institutional Review boards. Two groups of patients were included in the study. The first group included patients from Yale-New Haven Hospital (YNHH) who had EIF1AX mutation detected preoperatively, and thyroid surgery performed between March 2016 and May 2023. This group included patients with molecular testing performed on FNA samples with indeterminate cytology diagnosis which were re-classified according to the Bethesda System for Reporting Thyroid Cytopathology (TBSRTC) 3rd edition. Those included Atypia of undetermined significance (AUS) and follicular neoplasm (FN)/oncocytic follicular neoplasm (OFN), i.e. TBSRTC category III or IV diagnosis respectively [22]. In thyroidectomy specimens, we correlated preoperative ultrasound and FNA findings with final pathology which included identification of biopsy site changes in the target nodule. The second group included patients from MSKCC with carcinoma of thyroid origin and EIF1AX mutation detected, on primary thyroid tumor, local recurrence, or distant metastasis, diagnosed between 2010 and 2020. The diagnosis was rendered on a core biopsy, surgical resection, or FNA cytology which was also used to perform molecular analysis. Patients’ demographics, cytology and surgical diagnosis, molecular results and clinical follow-up were recorded.

All surgical pathology slides were reviewed by 3 pathologists (RA and MLP at YNHH, BX at MSKCC) who were blinded to any additional molecular alterations. The original histopathologic diagnosis and re-classification according to the WHO 5th ed. were recorded for all cases [21]. Cases with multifocal benign follicular-derived nodules with variable architecture, consisting of an admixture of large and small sized follicles, sometimes with papillae formations, were classified as Follicular nodular disease (FND) after identifying the target nodule as described above. Cases with a single or predominant totally encapsulated follicular-patterned tumor distinct from the background were classified as Follicular Adenomas (FA). FAs and the target nodule in FND were reported as having predominantly microfollicular, macrofollicular or mixed growth pattern (Supplemental table). Cases with a single or predominant totally encapsulated follicular cell derived neoplasm composed of > 75% oncocytic cells were classified as Oncocytic Adenomas (OA). Follicular-derived thyroid carcinomas with high grade features, defined as increased mitotic count and tumor necrosis, were classified as high grade follicular cell-derived non-anaplastic carcinomas (HGC). Those included poorly differentiated thyroid carcinomas (PDTC), diagnosed using the Turin proposal (solid, trabecular or insular growth pattern, absence of nuclear features of papillary thyroid carcinoma, presence of at least one of the following 3 features: convoluted nuclei, ≥ 3 mitotic figures/2 mm2, tumor necrosis) [23] and high grade differentiated thyroid carcinomas (HGDTC) using the MSKCC criteria (≥ 5 mitotic figures/2 mm2 and/or tumor necrosis) [24]. Follicular thyroid carcinoma (FTC) and oncocytic (Hurthle-cell) carcinoma (OCA) were defined as invasive malignant well-differentiated follicular neoplasm and invasive malignant well-differentiated follicular neoplasm composed of at least 75% of oncocytic cells respectively, without nuclear features of PTC or high grade features. Angioinvasive FTC (A-FTC) was defined as FTC with invasion of vessels within the tumor capsule or beyond. Molecular alterations were correlated with the cytologic and histologic diagnosis of the target nodule.

Molecular analysis

Molecular analysis on YNHH specimens was performed preoperatively on FNA cytology by next-generation sequencing (NGS)-based ThyroSeq version 2 (v2) (n = 4) or ThyroSeq version 3 (v3) (n = 10) at the University of Pittsburgh Medical Center (UPMC), or ThyroSure (n = 8) at YNHH. ThyroSeq v2 is a NGS assay assessing 56 thyroid-related genes for single nucleotide variants (SNVs), small insertions/deletion (indels), gene fusions and gene expression analysis. ThyroSeq v3 was expanded on its previous version to include 112 genes which cover additional genetic alterations and copy number variations (CNVs). It also uses a genomic classifier to separate malignant from benign lesions [25, 26]. ThyroSure, a modified thyroid genomic classifier, is a NGS assay developed at the Yale Pathology Molecular Diagnostics Laboratory, which is certified under the Clinical Laboratory Improvement Amendments (CLIA) Act of 1988 to perform high complexity clinical laboratory testing. ThyroSure is performed on extracted DNA and RNA from FNA samples and provides analysis of 78 thyroid-related genes to detect, SNVs, indels, gene fusions, gene expression alterations and also uses a genomic classifier to further stratify the cancer risk. Specimens from MSKCC (32 surgical and 2 cytology specimens) were tested by MSK-IMPACT at MSKCC, a deep-coverage targeted NGS technique detecting SNVs, indels, CNVs, and fusion/structural variants in 368 to 505 cancer-related genes, using custom DNA probes designed for targeted sequencing of all exons and selected introns, including canonical and selected non-canonical transcripts [27, 28]. All platforms included testing for EIF1AX mutations in exons 2, 5 and 6.

Results

YNHH patients

There were 916 TBSRTC category III and IV cytology specimens with molecular testing available for the study period and EIF1AX mutation was present in 48 cases, resulting in a mutation prevalence of 5.2% at YNHH. Of these, 22 patients [17 (77%) women, 5 (23%) men, median age 66 (range, 44 – 81) years] (Table 1) underwent surgical resection. Cytology diagnosis included AUS in 15 cases, FN in 4 cases and OFN in 3 cases. 8 cases had EIF1AX mutation only and 14 cases had at least one additional molecular alteration. Thyroidectomy specimens were reviewed, and the histopathologic diagnosis of the target nodule was rendered according to the new WHO Classification of Tumors 5th ed [21]. There were 9 FNDs, 3 FAs, 4 OAs, one A-FTC, one OCA and 2 HGCs (one high grade FTC and one PDTC). Two cases with a follicular growth pattern, crowding, nuclear enlargement with clearing and scattered grooves were classified as Non-invasive follicular thyroid neoplasms with papillary-like nuclear features (NIFTP) (Table 1, Fig. 1). In cases classified as adenomas and NIFTPs, the entire thyroid lobe (n = 2), nodule (n = 1) or capsule (n = 1) had originally been entirely submitted.

Table 1 Clinicopathologic Data (YNHH patients)
Full size table
Fig. 1
figure 1

Histology of representative cases of thyroid lesions classified according to WHO 5.th ed. A- Non-invasive follicular thyroid neoplasm with papillarylike nuclear features (NIFTP) with EIF1AX and KRAS mutations (H&E, X10), inset showing nuclear enlargement, irregular nuclear membranes and powdery chromatin (H&E, X800), B- EIF1AX mutated oncocytic carcinoma with lymphovascular invasion (H&E, X400), C- Follicular nodular disease with EIF1AX and NRAS mutations (H&E, X400), D- Poorly differentiated carcinoma with insular pattern harboring EIF1AX and NRAS mutations (H&E, X200)

Full size image

In FNA cytology samples with TBSRTC category III or IV diagnosis and EIF1AX mutation, the overall ROM in the YNHH cohort was 18% (4/22). Eight cases were positive for isolated EIF1AX mutation: 4 (50%) were FND, 3 (38%) were FA or OA and one (12%) was an OCA yielding a ROM of 1/8 (12%). There were 12 cases with EIF1AX mutation and one additional molecular alteration, 5 (42%) of which were FND, 4 (33%) were FA or OA, 2 (17%) were NIFTP and one (8%) was a PDTC yielding a risk of NIFTP of 17% and a ROM of 8%. There were 2 cases with EIF1AX and 2 additional molecular alterations, both of which were FTCs including one high grade FTC conferring a ROM of 100% (Tables 2 and 3). The ROM was 21% in cases with EIF1AX mutation and at least one additional molecular alteration. EIF1AX mutation consisted of a splice site mutation in 77% (17/22) and missense mutation in 23% (5/22) of the cases. All 8 cases with isolated EIF1AX mutations were of the splice site type with a ROM of 12%. All cases with EIF1AX missense mutation had additional molecular alterations present and those consisted of CNV in FND with oncocytic changes, TSHR in FND and OA, NRAS in PDTC, TERT and TP53 in high grade FTC (Supplemental table).

Table 2 Correlation of Molecular Alterations with Surgical Follow-up (YNHH patients)
Full size table
Table 3 Correlation of Molecular Alterations with Histopathologic Diagnoses (YNHH and MSKCC)
Full size table

Of 16 benign nodules (FND or adenoma), 7 (44%) had an isolated EIF1AX mutation, and 9 (56%) had one additional molecular alteration (Table 2), including CNV, GE, NRAS, HRAS, TSHR, TERT promoter and TP53 mutations (Supplemental table). All but 2 of the FND cases were EIF1AX splice site mutation (Table 4). NIFTP cases (n = 2) carried an EIF1AX mutation and a RAS mutation each. There were 4 malignant cases, 2 of which had 2 molecular alterations in addition to EIF1AX mutation: a widely invasive high grade FTC with combined TERT promoter, TP53 and EIF1AX missense mutations, and an encapsulated FTC with extensive angioinvasion and combined TERT promoter, HRAS and EIF1AX splice mutation. The other 2 malignant tumors were a PDTC with NRAS and EIF1AX missense mutation and an OCA case with an isolated EIF1AX splice mutation (Supplemental table).

Table 4 Type of EIF1AX Mutation by Histopathologic Diagnosis
Full size table

MSKCC patients

Thirty-four patients with thyroid tumors showing EIF1AX mutation were included [17 (50%) women, 17 (50%) men, median age 67 (range 43–85) years]. Testing by MSK-IMPACT was performed on 21 tumors from the primary tumor site and 13 local recurrence/regional recurrence/distant metastasis. All tumors were malignant, the histopathologic subtype was ATC in 14 cases, HGC in 12 cases, A-FTC in 2 cases and differentiated thyroid carcinoma (DTC) in 6 cases (differentiated thyroid carcinoma – not otherwise specified: DTC-NOS; n = 3, tall cell variant of papillary thyroid carcinoma: PTC-TCV; n = 1, follicular variant of papillary thyroid carcinoma: PTC-FV; n = 2). All 3 cases classified as DTC-NOS were follicular-patterned distant metastasis (bone, kidney, liver) without high grade features diagnosed on core biopsies. The nature of core biopsies prevented definite classification. All MSKCC cases harbored EIF1AX mutation with additional molecular alterations, and those included NRAS (n = 21), HRAS (n = 4), KRAS (n = 2), TERT promoter (n = 25) and TP53 (n = 13) mutations. Other molecular alterations included ATM, AXIN2, BBC3, BRAF, BREBBP, CUL1-EZH2, CDKN2A, CDKN2C, DAXX, GNAS, MLL1, MLL2, NF1, PAK7, PBRM1, PIK3CA, PTPRT, RAC1, SMAD2, TET2, TSHR, DDR2, PRKAR1A, RAD21, RBM10, SDHA, SDHB, TP53BP1, JAK2, NEGR1, SF3B1, PTCH1, TGFBR1, KLF4, TGFBR2, ZFHX3, NOTCH1 rearrangement, CCNE1 amp, PIK3CD del, CD79B gain and loss of FANCC (Supplemental table). EIF1AX mutation consisted of a splice site mutation in 26 and missense mutation in 8 cases (Table 4). TERT promoter mutation occurred in 64% (9/14) of ATCs, 83% (10/12) of HGCs, 100% (2/2) A-FTCs and 67% (4/6) of DTCs. TP53 mutation occurred in 79% (11/14) of ATCs, 0% (0/12) of HGCs, 50% (1/2) of A-FTCs and 17% (1/6) of DTCs. All 34 cases had at least 2 molecular alterations in addition to EIF1AX mutation including TERT promoter and/or TP53 mutations in 29 (85%) ATC and 22 (85%) HGC. RAS mutation was present in 27/34 (79%) cases (Table 3 and Supplemental table).

Discussion

Our study confirms previous reports that EIF1AX mutations occur in both benign and malignant thyroid nodules. They were originally reported to confer a ROM of 20% approximately [10]. However, subsequent studies revealed different ROM estimates depending on the type of EIF1AX mutation and the presence of additional molecular alterations [17, 19, 29]. In cases with indeterminate cytology and isolated EIF1AX mutation, the ROM ranged from 13% to 47.6% but higher ROM/NIFTPs, ranging between 70 and 80%, have been reported in cases with EIF1AX mutation co-existing with other driver mutations, reaching 100% in cases with EIF1AX splice site mutation and one additional molecular alteration for instance [10, 17,18,19, 29]. The findings suggest that the presence of an additional molecular alteration could represent a step into a malignant progression.

Most tumors with isolated EIF1AX mutation are well-differentiated carcinomas or NIFTPs, but co-occurrence of EIF1AX and RAS mutations correlates with larger tumors, aggressive behavior, advanced disease and predicts for shorter survival [4, 9, 10, 12, 13, 20]. Furthermore, the presence of TERT promoter or TP53 mutation with EIF1AX mutation confers a 100% ROM with a higher risk of a more aggressive malignancy such as FTC and OCA but has mostly been associated with PDTC and ATC [4, 5, 12, 14, 17,18,19, 29, 30]. Indeed, PDTCs and ATCs are characterized by distinct genomic profiles with multiple molecular alterations; although BRAF V600E and RAS mutations are the main drivers, PDTCs and ATCs were also reported to be enriched for EIF1AX mutations, frequently associated with mutations in the TERT promoter, TP53 or genes encoding PI3K/AKT/mTOR pathway effectors or chromatin modifiers, which are major drivers of tumor progression [4, 10, 12, 31,32,33]. In a series of PDTCs and ATCs with concurrent well differentiated PTC and/or nodular hyperplasia components, Simões-Pereira et al. reported a case of PDTC case with co-existing EIF1AX and RAS mutations in the PDTC component, isolated RAS mutation in the well differentiated PTC component and no molecular alteration detected in the nodular hyperplasia component, suggesting the dedifferentiation may be driven by accumulation of multiple genetic events [30].

Our results are consistent with the literature. In the general population at our tertiary care center, the overall ROM was 18% and this risk increased when multiple molecular alterations were present, with TP53 and/or TERT mutations being associated with aggressive phenotypes. To reflect daily practice, we have limited our YNHH study cohort to pre-operative EIF1AX mutation detected in TBSRTC category III and IV thyroid cytology cases where molecular testing is performed [17,18,19, 29]. We should however note the limitation of including the indeterminate cytology samples only for the purpose of assessing the relationship of EIF1AX mutation with malignancy, since TBSRTC category III and IV carry a high prevalence of RAS mutations in general, which are mutations encountered in both benign, low-risk and malignant neoplasms, and that the ROM in those categories is only 22% and 30% respectively [22]. We therefore included a separate set of patients with known malignancy from a cancer referral center. All malignant cases from MSKCC harbored at least 2 molecular alterations in addition to EIF1AX mutation and were associated with aggressive clinicopathologic characteristics, supporting that co-occurrence of multiple molecular alterations with EIF1AX mutation is associated with an aggressive phenotype and contributes to an unfavorable clinical course.

The WHO 5th ed. has defined multifocal benign proliferation of thyroid follicular cells resulting in multiple clonal and non-clonal nodules with variable architecture as FND and discouraged the use of the previous designation of adenomatoid hyperplasia, goiter, diffuse goiter or colloid nodule suggesting that clonal nodules represent true neoplasms [21]. Molecular alterations identified in the clonal nodules include genetic variants of a few genes including RGS12, GRPEL1, CLIC6 and WSF1 in familial goiters [34] and somatic alterations including SPOP, EZH1 and ZNF148 genes [35]. Thus, with the new classification, FND falls under benign tumors [21, 36,37,38,39]. It is possible that the presence of EIF1AX mutation in both NIFTPs and adenomas, as well as in the FND cases in our study represents the molecular alteration in a clonal nodule, and the “hyperplastic” nodules harboring EIF1AX mutation, reported in many studies, are probably clonal nodules and should be designated FND. In this setting, an isolated EIF1AX mutation in nodules comprising FND could represent an early genetic event in the multistep process of thyroid neoplasia. It would not be sufficient for malignant transformation when occurring in isolation and requires other mutations for progression to overt malignancy. In fact, the progressive accumulation of molecular alterations from benign to malignant tumors, with suggestions that FAs may be precursors to FTCs are known [40]. As evidenced by other studies [10] and supported by ours, the fact that EIF1AX is encountered as the sole molecular alteration in benign neoplasms supports this theory. There was one OCA case with an isolated EIF1AX mutation in this study; EIF1AX mutation might have occurred as a late event in this case or it might have been the driver mutation with an additional undetected molecular alteration that could have been responsible for the malignant process.

Similar to the results of the TCGA and a few other studies, our study shows that isolated EIF1AX mutation was most likely encountered in follicular and oncocytic type neoplasms such as FA, OA and OCA. EIF1AX mutated tumors, with or without RAS mutation, were reportedly PTCs with a follicular phenotype, typically encapsulated, as well as FAs, OAs, FTCs and OCAs [5, 9,10,11, 14, 18, 29, 41,42,43]. Contrary to other studies, there were no PTCs with isolated EIF1AX mutation in our study.

EIF1AX mutations appear to cluster into different regions of the gene depending on the tumor type [6, 7]. In thyroid carcinomas, all mutations are single nucleotide substitutions that are clustered in two specific areas of the gene: either in codons 6–15 near the N-terminal domain in exon 2 as a missense mutation as seen in uveal melanomas [6] or more commonly, in codon 113 at a hotspot splice acceptor site between exons 5 and 6 in the C-terminal domain of EIF1AX (X113_splice mutation) resulting in a 12 amino acid deletion [10, 12]. EIF1AX splice mutation is specific to thyroid cancer and confers a higher ROM compared to exon 2 mutation, but this finding has not been universally confirmed with both mutation types being reported in malignant and benign thyroid lesions [9, 10, 14, 17,18,19,20]. We have found similar results; both EIF1AX splice site and missense mutations were encountered in carcinomas, adenomas and FND; EIF1AX splice site mutation was associated with a ROM of 12% while all missense mutated tumors had at least one molecular alteration in addition to EIF1AX mutation. It is unclear whether EIF1AX missense mutation was the driver gene in these cases, since many of the additional molecular alterations have been associated with thyroid pathogenesis [4, 12, 31, 44]. There were no cases of isolated EIF1AX missense mutation in our study, limiting the assessment of its clinical significance.

There are multiple drawbacks to our study inherent to its retrospective nature and the selection bias at the two institutions. The use of multiple molecular testing platforms represents one shortcoming of this study. While NGS panels are relatively sensitive, the difference in sensitivity among the different platforms may alter the detection of EIF1AX and other molecular alterations which could be missed if a lower sensitivity platform is used. Furthermore, we have documented the presence of EIF1AX mutation in FND and various thyroid lesions in the general population, but its true prevalence in each of those categories is difficult to assess. Only a subset of patients with indeterminate cytology diagnosis, probably those with concerning clinical and radiologic features, underwent surgery. Additionally, molecular studies are not performed on Bethesda Category II. Therefore, the prevalence of EIF1AX mutation in FND remains unknown, but is likely to be underestimated. Similarly, the assessment and distribution of EIF1AX mutation in malignant thyroid tumors is largely unknown since no category V or VI were included in the YNHH patients. To review the histology and molecular findings of EIF1AX-mutated aggressive cancers, we included a set of patients with known thyroid carcinomas from a cancer referral center. While the inclusion of patients with advanced disease reflects a selection bias towards aggressive malignancies characteristic of referral practice, it nevertheless reveals a different distribution of molecular alterations in EIF1AX-mutated cancers compared to EIF1AX-mutated nodules with indeterminate cytology. In addition, the pathologic and molecular characteristics of the tumors in this group confirm the findings from the general population. It is important to stress that this study does not investigate the prevalence of EIF1AX in thyroid or investigate its distribution across thyroid lesions. It rather shows that EIF1AX mutations are present in thyroid carcinomas as well as benign nodules including FND and reflects the interpretation of EIF1AX mutation in thyroid FNA specimens classified as Bethesda category III or IV as encountered in daily practice.

In conclusion, EIF1AX mutation is present in benign and malignant thyroid nodules. An isolated EIF1AX mutation detected pre-operatively in Bethesda categories III and IV confers a low ROM while the accumulation of additional molecular alterations is associated with increased ROM and aggressive forms of thyroid cancers.