Thymic epithelial tumors (TETs) are uncommon malignant neoplasms with an incidence of approximately 0.32 per 100,000 people/year (1). According to the 2020 classification of the World Health Organization, TETs are divided into thymomas and thymic carcinomas. Thymomas account for about 90% of TETs and are further classified into A, AB, B1, B2, and B3 histotypes depending on their morphological features (2). About 40% of thymomas show myasthenia gravis, a paraneoplastic autoimmune syndrome, which is absent in thymic carcinomas and micronodular thymomas (3). Myasthenia gravis is more common in type B than type A and AB thymomas (4). Historically, tumor stage has been defined according to the Masaoka and Koga system, but more recently, it is determined according to the eighth edition of TNM (5). Surgery is the cornerstone treatment for localized disease, ensuring definitive tumor eradication in most stage I and II TETs (6). Chemotherapy with schedules containing cisplatin and anthracyclines is effective in TETs and can be used in the neoadjuvant setting to bring back to resectability locally advanced stage III tumors (7). While hematogenous or lymphatic metastases are common in thymic carcinomas, thymomas frequently grow through local infiltration, and their most frequent way of diffusion is through pleural metastases (2). Several phase II trials investigated the efficacy of chemotherapy for the first-line treatment of metastatic TETs, showing an objective response in 50-90% of cases (8). Unfortunately, after disease progression, further lines of chemotherapy become much less effective, and there is an urgent need for novel treatments. Even if GTF2I mutations are common in A and AB thymomas, there are no clinically relevant targets in TETs except for KIT mutations observed in only 10% of thymic carcinomas (9). Nevertheless, targeted therapies such as sunitinib and lenvatinib have shown good efficacy in pretreated thymic carcinomas, with an objective response rate of 26% and 38%, respectively (10, 11). Immunotherapy is effective in TETs but severely increases the risk of autoimmune-mediated complications, especially in thymomas, and therefore can be considered only in some cases of thymic carcinomas (12, 13).

Several attempts to identify somatic mutations that characterize TETs have been made in recent years, including whole-genome sequencing of two cases, exome sequencing both from our group and from The Cancer Genome Atlas (TCGA), and targeted resequencing using panels of genes frequently mutated in other kinds of cancers (4, 9, 14–21). While somatic mutations are usually observed in thymic carcinomas, thymomas present only occasional mutations, with the exception of GTF2I, which is common in A and AB thymomas. Reviewing the literature, we defined a panel of genes frequently mutated in TETs and performed targeted resequencing of a series of surgically resected thymomas to favor the identification of mutations in this subset of tumors.

Reviewing the results of whole genome, exome and targeted sequencing of TETs, we identified 4208 mutations in 339 patients and defined a panel of 63 genes frequently mutated in these tumors. We designed a targeted resequencing assay using Haloplex technology from Agilent and sequenced 67 thymomas resected from patients with myasthenia gravis. The only recurrent mutation observed in our analysis and in the reviewed literature was the missense L424H GTF2I mutation. There were no other oncogenes with recurrent mutations, while several tumor suppressor genes were predicted to be inactivated by the identified mutations.

TETs exhibit divergent clinical behavior: while some tumors grow as an expansive mass, others infiltrate the mediastinal fat and the surrounding structures or metastasize mostly into the pleural cavity (2). Moreover, some thymomas are associated with paraneoplastic syndromes, including myasthenia gravis (40%), Aplastic Anemia, and Hypogammaglobulinemia This association is likely due to impaired negative selection of thymocytes that recognize self-antigens in the aberrant thymus (2). Somatic mutations, copy number aberrations, and mRNA expression correlate with different kinds of TETs and help to differentiate tumors with different clinical features (9). TETs have one of the lowest tumor mutation burden observed in solid tumors, with an average of 0.48 mutations per mega-base (4). However, thymic carcinomas have a significantly higher number of mutations compared to thymomas (9). The most common mutation observed is the missense L424H mutation in GTF2I, which is unique to this type of neoplasm (38). Therefore, we decided to design a panel of genes for targeted resequencing specific to TETs. In 2018, we reviewed the literature identifying 4208 somatic mutations in 339 TETs, selecting those genes with recurrent mutations. Since then, additional reports of targeted resequencing of TETs evaluating common cancer genes have been published, confirming previous results (18, 21, 39, 40). Consequently, our current analysis remains updated to define the most relevant mutations in thymomas. The low mutation burden detected in thymomas has been attributed to the low number of somatic mutations in the DNA of tumor cells. However, this may be underestimated in some tumors where non-tumoral thymocytes outnumber cancer cells. This is frequent in AB, B1 and B2 histotypes where the number of tumor cells could be less than 10% of those present in the sample (2, 41). It is not surprising that mutations cannot be identified in some thymomas. Targeted resequencing can help identify some of these mutations, especially when a deep coverage is obtained. A targeted resequencing approach limited to the genes of interest can help achieve deep coverage, which is useful for detecting mutations in samples rich in non-neoplastic cells. Using the Halopex custom panel, we obtained a mean coverage of 1248 reads (average of the mean coverage of all samples; range 353-2542). However, after confirming the data with the Illumina Enrichment custom kit, we noticed that mutation calls with an allele fraction below 4% and a minimum coverage of 20 reads had a poor concordance between the two platforms. This could represent a limitation in detecting somatic mutations in samples rich in thymocytes and non-neoplastic cells of tumor microenvironment.

GTF2I mutation was the most common, present in 21% of the analyzed TETs, with 50% found in A thymomas and 38% in AB thymomas. This appears to be an underestimation compared to previous findings of 82% and 100% in A thymomas and 74% and 70% in AB thymomas reported by Petrini I et al. (9) and Radovich M. et al. (4), respectively. In the present evaluation, only patients with myasthenia gravis were included, which may introduce a selection bias compared to previous reports. Indeed, thymoma-associated myasthenia gravis is more common in B thymomas than A and AB histotypes and is absent in thymic carcinomas and micronodular thymomas (2, 4, 42). Moreover, in the present evaluation, tumor specimens were sampled from the resected mass and immediately frozen, potentially increasing the detection rate of GTF2I mutations using highly sensitive technologies. Using PyClone analysis, the presence of GTF2I mutation was observed in up to 64% (14/22) of thymomas including in those with B histotype (43%; 6/14) (43).

The missense mutation of GTF2I on chromosome 7:74146970 leads to a leucine-to-histidine substitution in the second conserved TFII-I repeat domain of the protein, near to the DNA binding site. There are six known isoforms of GTF2I, but thymomas express only isoforms 2 (Beta) and 4 (Delta), according to RNA sequencing results (9). L424H mutated GTF2I is an oncogene. The CRISPR/Cas9 knock-in of Gtf2i mutation leads to neoplastic transformation of murine immortalized thymic epithelial cells, which acquire the ability to form tumors when transplanted under the skin of nude mice (44). Moreover, two different mouse models expressing mutated Gtf2i in the thymus have been generated by two independent groups (45, 46). He Y. et al. generated a mouse model in which the expression of wild type Gtf2i is switched to L424H mutant, in the presence of Cre recombinase under the control of its own promoter. Subsequently, Gtf2i L424H-floxed mice were crossed with Foxn1-Cre mice to obtain Gtf2i L424H conditional KI mice, in which mutant Gtf2i L424H is expressed in Foxn1-expressing TECs (46). Giorgetti O.B. et al. generated a transgenic mouse model by inoculating a vector containing the mutant delta Gtf2i isoform under Foxn1 promoter into pronuclei of fertilized mouse eggs (45). In both cases, young mice showed defects of thymic medulla development and maturation of medullary epithelial cells, while aged mice developed thymomas. Histologically, thymomas mirror human type B1 and B2 histotypes. The evaluation of the transcriptome of thymoma cells in transgenic mice showed a dysregulated expression of the genes associated with cortical, medullary and progenitor differentiation. According to TCGA data, human TETs with the GTF2I L424H mutation had an enrichment in cortical and intertypical thymic epithelial cell signatures compared with the GTF2I WT tumors (Supplementary Figure S2), which mirrors thymomas in the Gtf2i mutated mouse model. Differences between the mouse model and human tumors are expected. In human tumors, the somatic mutation of GTF2I is acquired in adult cells, whereas in mouse models, the mutation is already expressed during embryonal development of the thymus. In normal mice, Gtf2i and Foxn1 follow a similar pattern of expression. Gtf2i and Foxn1 are highly expressed in early precursor cells and are subsequently more expressed in cortical than in medullary epithelial cells (45). Therefore, it is not surprising that transgenic mouse models expressing mutated Gtf2i under the promoter of Foxn1 develop thymomas with cortical features instead of medullary features. Experiments using murine epithelial cells transformed by the CRISPER/Cas9 knock-in of Gtf2i mutation suggest enhanced resistance to apoptosis, increased resistance to DNA damage, and the ability to grow under stress conditions such as glucose deprivation (44). However, it remains to be elucidated how GTF2I mutation drives the neoplastic transformation of thymic epithelial cells suggesting that these novel murine models harboring Gtf2i L424H mutation represent an ideal tool for further studies.

Copy number gain and loss of 7q11.23, a cytoband containing the locus of GTF2I, are associated with Williams-Beuren and with 7q-microduplication syndromes two conditions with a symmetrically opposite phenotype. (PMID: 25501393) None of the patients with thymoma included in our evaluation had the diagnosis or symptoms suggesting these conditions.

Reviewing mutations described in the literature, we observed at least two groups of TETs with mutually exclusive mutations: one with GTF2I and another with TP53 mutations. GTF2I mutations co-occurred with those of HRAS, TTN, and UNC93B1 in the same group of tumors. On the contrary, mutations of BRCA2, SETD2, PBRM1, and CDKN2A significantly co-occurred with TP53 mutation. In our series of thymomas, we did not observe the division in these two groups, probably because of the absence of thymic carcinomas. In our cases of thymomas from patients with myasthenia gravis, we identified a trend for the presence of mutually exclusive mutations of GTF2I and BCOR. BCOR is a tumor suppressor gene that encodes for an epigenetic regulator involved in the specification of cell differentiation and body structure development and takes part in the non-canonical polycomb repressive complex 1 (47). BCOR inactivating mutations have been previously described in thymomas (17) and in other kinds of tumors, including sarcomas (47) and lymphomas (48). In TCGA report, integrated analysis of CNV, mRNA, miRNA, DNA methylation, and reverse phase protein array identified four clusters of TETs (4). Subtype 1 is primarily represented by type B, subtype 2 by type TC, subtype 3 is primarily type AB, and subtype 4 is a mix of types A and AB. As expected, subtype 1 (mostly type B) was greatly enriched for cases with myasthenia gravis. Cases in subtypes 1 and 3 were associated with higher lymphocyte content, whereas GTF2I mutation was predominantly seen in subtypes 3 and 4. It remains to be understood if the presence of these clusters of gene expression is sustained by just two or more pattern of genomic mutations, whose detection remains challenging in tumors rich of lymphocytes. Enrichment of tumor cells in combination with deep coverage of sequencing could contribute to elucidate the somatic mutations of thymomas rich in non-neoplastic thymocytes.

In the current evaluation, C to A transversions were more common, whereas, according to the literature, C to T transitions were more frequently observed in TETs. Different mutation patterns have been reported for different types of tumors. For example, lung cancers share a mutation spectrum dominated by C to A mutations, consistent with their exposure to the polycyclic aromatic hydrocarbons in tobacco smoke, whereas melanomas show frequent C to T mutations caused by an altered repair of ultraviolet-induced covalent bonds between adjacent pyrimidines (49). Since mutations are more common in thymic carcinomas, the high incidence of C to T transitions reported in the literature could be related to these kinds of TETs. On the contrary, our analysis includes only thymomas with myasthenia gravis and possibly a different pattern of mutations sustained by a different etiological factor can be supposed for these tumors. In the literature, COSIMIC 1 mutational signature (spontaneous deamination of 5-methylcysteine) was observed in GTF2I mutant and wild-type tumors. On the contrary, COSMIC 5 (unknown etiology) was observed in GTF2I tumors and COSMIC 6 (defective mismatch repair) in TP53 mutated tumors. In the current series, the presence of COSMIC 1 signature and of a signature associated to mismatch repair defect was observed (COSIMC 15). COSMIC 1 signature is associated to spontaneous deamination of 5-methylcysteine, a phenomenon observed in aging. This observation is concordant with the biology of TETs that usually occurs in older people. Mutational signatures could suggest a similar etiology for the initial transformation of thymic epithelial cells, but subsequent different routes of mutations in GFT2I mutated and in TP3 mutated TETs.

Several targeted therapies have been tested for the treatment of advanced TETs with limited clinical efficacy. GTF2I was the only recurrent mutation between the literature and the current sequencing analysis. To date, there is no drug able to inhibit mutated GTF2I. Moreover, GTF2I mutated tumors usually present an indolent clinical behavior and surgery is curative for most of the cases. In order to identify possible targets for therapies, we used the M-CAP score to predict the effect of the observed mutations on oncogenes and tumor suppressor gene function. M-CAP is a classifier of somatic mutations that combines the results of multiple algorithms in a score with high sensitivity in predicting the chance of a functional or a structural aberration of the protein. Among oncogenes, HRAS A146V, ROS1 W969L, KIT R588M, JAK3 S680R, RET T676M, SF3B1 I704T, FGFR3 S269C and GLI R864W had significant M-CAP scores. ROS1, RET, FGFR3 and KIT are tyrosine kinase receptors of growth factors and JAK3 is a tyrosine kinase associated to growth factor receptors. Different inhibitors are available for each one of these genes, but the episodical observation in few samples limits our chance to design a clinical trial for a specific target. Moreover, the mutation should confer a gain of function of the protein in order to transform a proto-oncogene into an oncogene. None of these tyrosine kinase mutations has been reported to be oncogenic in the literature. The mutation of FGFR3 S269C can activate the receptor by adding a cysteine residue in its extracellular domain implicated with altered skeletal formation when germline but not with the formation of tumors. FGFR3 is considered an oncogene but has also tumor-suppressive properties in cells with epithelial phenotype (50). Only HRAS A146V has been previously described in colorectal cancers, and it is possibly implicated in tumorigenesis (51). However, the HRAS mutation observed is not the classical mutation of codon G12 or G13, which confers a slower hydrolysis of GTP. Specific inhibitors are available for the G12C mutation (52). Interestingly, the HRAS A146V co-occurred with GTF2I L424H mutation as predicted by the concordance table of the mutations observed in the literature. Tumor suppressor genes can be inactivated by several different mutations and multiple genes. We observed three inactivating mutations of NF1, supporting the importance of RAS activation for the development of TETs. NF1 functions as GAP protein, enhancing the hydrolysis of GTP bounds to RAS to GDP and switching off the downstream transduction of the intracellular signal pathway. Targeted therapies for inactivated tumor suppressor genes are not available to date and specific targets for aggressive thymomas remain to be identified for effective treatment.

From a clinical perspective, our analysis did not identify any mutations immediately eligible for targeted treatment. However, it does suggest a molecular subclassification of thymomas into at least two subgroups. We have confirmed that GTF2I mutated tumors exhibit a better prognosis because diagnosed only in stage I-II. Conversely, the other group of thymomas demonstrates a worse prognosis, and although a highly prevalent driver mutation has not been identified yet, BCOR mutations appear to be enriched within this subgroup. Due to the low incidence of thymomas, single-arm phase II trials are typically conducted to assess the efficacy of novel drugs. In order to better contextualize the results of these trials, it is important to include a prediction of the clinical behavior of thymomas based on their molecular aberrations. Our designed panel of genes could serve as a useful tool for the molecular classification of thymic epithelial tumors. This is particularly valuable because some of the most interesting genes mutated in thymomas are not included in standard panels for next-generation sequencing.

In conclusion, based on a comprehensive review of the literature, we have developed a gene panel for targeted resequencing of thymic epithelial tumors and confirmed the presence of GTF2I mutations in a subgroup of thymomas among patients with myasthenia gravis.

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

The studies involving human participants were reviewed and approved by Ethics Committee of Tuscany Region for Clinical Trials - Section of North West Area (CEAVNO). The patients/participants provided their written informed consent to participate in this study.

EP: Methodology, Investigation; FCu: Software, Investigation, Writing - Review & Editing; SP: Writing - Review & Editing; GT: Methodology; AV: Investigation; FCo: Resources, Writing - Review & Editing, Funding acquisition; VN: Resources; MG: Investigation, Writing - Review & Editing; MM: Resources; RR: Supervision; VA: Resources; MA: Resources; SB: Investigation, ML: Resources, Supervision; IP: Methodology, Writing - Original Draft, Supervision, Funding acquisition. All authors contributed to the article and approved the submitted version.