Landscape of gene fusions in epithelial cancers: seq and ye shall find

Some gene fusions are associated with distinct subtypes of carcinoma, while others
have been detected across different tissues or lineages, defining molecular subsets
of cancers transcending morphological distinctions.

Recurrent gene fusions as biomarkers of subtypes of solid cancers

Some of the salient gene fusions that define molecular subtypes of epithelial cancers
within specific organs or tissue types are summarized in Table 1. The ETV6–NTRK3 fusion is a diagnostic biomarker of secretory breast carcinoma, as well as the acinic
cell carcinoma or cystadenocarcinoma recently designated as “mammary analog secretory
carcinoma of salivary glands” (MASC) 21], 103]. BRD-NUT fusions define NUT midline carcinoma 104], 105]. CRTC–MAML2 fusions are the defining molecular aberration of mucoepidermoid carcinoma (MEC) 106], 107]; translocation-negative MECs are proposed to be designated as a distinct subgroup
of adenosquamous carcinoma 108]. CRTC-MAML fusions are also found in MEC of the lung 109]–112], cervix 113], thyroid glands and oral cavity 114], as well as in clear cell hidradenoma of the skin 115], 116]. In all cases, MAML2 fusions characterize benign or low-grade tumors, and for reasons not described so
far have been associated with a favorable prognosis 117]. Interestingly, pulmonary MECs have shown clinical response to gefitinib in the absence
of sensitizing EGFR mutations, suggesting a potential connection with CRTC–MAML2 and the possibility of therapeutic application in other MECs harboring this fusion
110], 118]. The diagnostic subclass of adenoid cystic carcinomas, including salivary gland and
breast cancer, is characterized by MYB-NFIB gene fusions 119], 120]. Fusions defining subtypes within a cancer include RET and NTRK gene fusions in subsets of papillary thyroid carcinoma 121], while PAX8-PPAR? fusions characterize subsets of follicular thyroid carcinoma 22], 122]. ETS family gene fusions, primarily including ERG (and less frequently, ETV1, ETV4, ETV5 or FLI1), are found in approximately 50 % of prostate cancers, the most common fusion being
TMPRSS2-ERG. The EWSR1–ATF1 fusion found in hyalinizing clear cell carcinoma of the salivary glands, a rare and
indolent tumor, can potentially be used as a molecular marker of this subtype that
is histologically similar to the more aggressive MEC 123].

Gene fusions or fusion partners found across tissue types are common in solid cancers.
The EML4–ALK fusion, initially identified in lung cancer 9], 10] has since been reported in breast cancer 124], colorectal carcinomas 66], 124], and in pediatric renal medullary carcinoma that affects young African–Americans
with the sickle cell trait 125], 126]. Similarly, RET fusions, first characterized in thyroid cancer, are widely observed in lung cancers,
and the EWSR1–POU5F1 fusion was detected in two rare epithelial tumors, hidradenoma of the skin and MEC
of the salivary glands 127].

Gene fusions involving RAF kinase genes (BRAF, RAF1, CRAF) have been identified in low-grade tumors of the central nervous system (pilocytic
astrocytomas and other low-grade gliomas), gastric cancer, melanoma and prostate cancer.
RAF family fusions involve truncation of the N-terminal auto-inhibitory domain, thus
generating constitutively active RAF protein. Curiously, BRAF gene fusions in low-grade astrocytomas have been associated with a tendency to growth
arrest, conferring a less aggressive clinical phenotype and a better clinical outcome
75], 128]. Additionally, RAF family fusions have been defined across diverse solid cancers,
including prostate, gastric, and skin cancers 82], 83]. A screen for BRAF gene fusions in 20,573 solid tumors, using the FoundationOneâ„¢ targeted gene panel,
identified BRAF fusions involving 29 unique 5? fusion partners in 55 (0.3 %) cases across 12 different
tumor types, including 3 % (14/531) of melanomas, 2 % (15/701) of gliomas, 1.0 % (3/294)
of thyroid cancers, 0.3 % (3/1,062) of pancreatic carcinomas, 0.2 % (8/4,013) of non-small
cell lung cancers and 0.2 % (4/2,154) of colorectal cancers, as well as single cases
of head and neck cancer, prostate cancer, rectal adenocarcinoma, ovarian, uterine
endometrial, and mesothelioma 70].

Fusions involving FGFR tyrosine kinase family genes have also been observed across
diverse cancers 88]. The first FGFR fusion observed in epithelial cancers, FGFR1-PLAG1, was found in a subset of pleomorphic salivary gland adenomas, and involves FGFR1 as the 5? partner upstream of PLAG1, the known driver of salivary gland tumors 91]. Curiously, this fusion excludes the tyrosine kinase domain of FGFR. Fusions that
retain the tyrosine kinase domain of FGFR include FGFR3–TACC3 in glioblastoma 92], 129]. Subsequently, diverse FGFR fusions, all retaining the tyrosine kinase domain, have been observed in bladder,
lung, breast, thyroid, oral, and prostate cancers, involving FGFR1, 2, or 3 either as the 5? or 3? partners 88], 94].

Some gene fusions provide personalized therapeutic targets

In Additional file 2 we summarize recent clinical trials involving gene fusions in epithelial cancers.
The RET inhibitor vandetanib shows antiproliferative activity in RET-mutant medullary thyroid cancer (MTC) 130], and was recently approved by the US Food and Drug Administration for treatment of
metastatic MTC. Sensitivity to vandetanib was also observed in RET-fusion-positive papillary thyroid carcinoma 131] and lung cancer cells 68], 132]. Treatment with Pfizer’s kinase inhibitor crizotinib (PF02341066) led to a dramatic
clinical response in EML4–ALK-positive NSCLC patients 133], 134], as well as in one patient with an SLC34A2–ROS1-fusion-positive tumor 58]. Unfortunately, resistance is inevitably observed, owing to mutations in the kinase
domain 134], 135], or ALK gene fusion amplification, KIT amplification or increased auto-phosphorylation of EGFR136]. This is representative of the challenge of treating solid cancers and argues for
the development of combinatorial therapeutic approaches from the start rather than
sequentially, as is the practice currently. RAF or MEK inhibitors represent potential
precision therapeutic options for several solid cancers with the diverse RAF family
gene fusions described earlier. Several FGFR inhibitors currently in clinical trials
represent potential therapeutics for cancers harboring FGFR fusions across multiple
cancer types, including bladder cancer, prostate cancer, and others 88], 90], 94], 137]. The rare PIK3C family gene fusions in prostate cancer (for example, TBXLR1-PIK3CA and ACPP-PIK3CB) show overexpression of the PI3KC genes and may be sensitive to PIK3CA inhibitors 83].

For treatment of secretory breast carcinoma expressing the ETV6–NTRK3 fusion, therapeutic targeting of the downstream signaling axis of IGF1R, using the
IGIFR/INSR kinase inhibitors BMS-536924 and BMS-754807 that are currently in clinical
trials, was found to be effective 138]. Breast cancer cells expressing NOTCH fusion products that retain the ?-secretase cleavage site were sensitive to ?-secretase
inhibitor (GSI) in culture, and treatment with GSI reduced tumor growth in vivo 86]. On the other hand, breast cancer cells harboring NOTCH fusions that encode NICD independent of the ?-secretase cleavage site were insensitive
to GSI.

In a recent clinical sequencing study of 102 pediatric cancers, among 37 non-sarcoma
solid cancers, several functional gene fusions were identified, including TFE3 fusions in a colorectal cancer (SFPQ-TFE3) and renal cell cancer (ASPSCR1–TFE3) — both cases were treated with pazopanib, the latter displaying stable disease for
10 months 139].

Efforts to target several other gene fusions are underway. The newly developed bromodomain
inhibitors that have shown dramatic efficacy in hematological malignancies 140], 141] are now being tested in multiple clinical trials for NUT midline carcinoma characterized
by BRD3/4-NUT gene fusions, which represent a rare but highly aggressive class of tumors with no
effective treatment currently available 104]. Also, the R-spondin fusions observed in colorectal and prostate cancer may be sensitive
to Wnt pathway antagonist porcupine inhibitors 142].

Gene fusions involving ETS transcription factors have been utilized in diagnostic
applications. A non-invasive assay system has been developed based on the detection
of TMPRSS2–ERG fusion transcripts in urine samples from patients, which in combination with the
detection of urine PCA3 improved the performance of the multivariate Prostate Cancer Prevention Trial risk
calculator in predicting cancer on biopsy 143]. Detection of TMPRSS2–ERG in circulating tumor cells in therapy-naive patients and in castration-resistant
prostate cancer patients following treatment suggests potential applications in non-invasive
monitoring of the therapeutic response 144]. While therapeutic targeting of transcription factor oncogenes is intrinsically challenging,
on the basis of the interaction of ERG with the DNA repair enzyme PARP1 and DNA protein kinase DNA-PKc, use of PARP inhibitors
was shown to inhibit growth of TMPRSS2-ERG-positive prostate cancer xenografts 145]. Additionally, PARP inhibition was associated with radiosensitization of TMPRSS2–ERG-positive prostate cancer cells 146], 147]. These experimental leads point to possible therapeutic avenues targeting a prevalent
gene fusion in a common carcinoma.