Oncogenic FGFR Fusions Produce Centrosome and Cilia Defects by Ectopic Signaling

A single primary cilium projects from most vertebrate cells to guide cell fate decisions. A growing list of signaling molecules is found to function through cilia and control ciliogenesis, including the fibroblast growth factor receptors (FGFR). Aberrant FGFR activity produces abnormal cilia with deregulated signaling, which contributes to pathogenesis of the FGFR-mediated genetic disorders. FGFR lesions are also found in cancer, raising a possibility of cilia involvement in the neoplastic transformation and tumor progression. Here, we focus on FGFR gene fusions, and discuss the possible mechanisms by which they function as oncogenic drivers. We show that a substantial portion of the FGFR fusion partners are proteins associated with the centrosome cycle, including organization of the mitotic spindle and ciliogenesis. The functions of centrosome proteins are often lost with the gene fusion, leading to haploinsufficiency that induces cilia loss and deregulated cell division. We speculate that this complements the ectopic FGFR activity and drives the FGFR fusion cancers.


Primary Cilium and Its Role in Cancer Development
A majority of the vertebrate cells are capable of forming a primary cilium, a microtubulebased organelle that projects from the centrosome to integrate signaling pathways and mediate cell-to-cell communication. Mutations in genes that control cilia structure or function produce a growing list of diseases called ciliopathies. To this day, at least 35 ciliopathies exist, and more than 400 candidate proteins have been identified [1]. Virtually all annotated ciliopathies are genetic developmental disorders; however, function of cilia in the tissue homeostasis is also beginning to emerge [2].
During cell division, the centrosomes need to function in the mitotic apparatus. Therefore, the cilium is typically disassembled during mitosis, even though cilia rudiments may be preserved [3,4]. The presence of a primary cilium is, therefore, tightly coupled with the cell cycle. In the majority of the cilia-competent cells, the primary cilium is formed during the G0/G1 phase of the cell cycle and resorbs before the S phase [5,6]. Several mitotic kinases, Aurora A [7,8], polo-like kinase 1 (PLK1) [9] and NIMA-related kinase 2 (NEK2) [10], were shown to block assembly and induce disassembly of the primary cilium, and upregulated activity of these kinases is frequently found in cancer [11][12][13][14][15][16][17][18][19]. Inhibition of the cilia disassembly signaling using small chemical inhibitors restored ciliogenesis and suppressed tumor growth in cholangiocarcinoma [20] or chondrosarcoma [21].
It is mainly the loss of primary cilia, as well as of their regulatory function in cellular signaling and cell division, that has been associated with neoplastic transformation and tumor progression [22][23][24][25]. In glioblastoma, disruption of ciliogenesis was observed at all stages, starting at early tumor lesions [26]. In a mouse model of Kirsten rat sarcoma virus

FGFR Gene Fusions in Cancer
Deregulated FGFR signaling, mostly caused by increased FGFR activity, has been implicated mainly in tumor progression, through poorly understood mechanisms involving accelerated proliferation, resistance to apoptosis and enhanced angiogenesis [93, [149][150][151][152]. Among the 4853 tumor samples analyzed by next generation sequencing, a FGFR aberration was found in 7.1% of all cases [153]. The most frequent lesion was gene amplification, accounting for 66% of FGFR aberrations [153], and typically resulting in FGFR overexpression and increased activity [154][155][156][157][158]. FGFR mutations were less frequent, covering 26% of the identified aberrations [153]. More than 200 distinct FGFR point mutations have been identified in cancer, targeting the extracellular, transmembrane and kinase domains of all four FGFRs [133, [159][160][161]. The majority of the mutations lead to ligand-independent FGFR dimerization and increased pathway activity [162][163][164][165]. Interestingly, somatic mutations found in cancer frequently overlap with those causing developmental disorders (extensively reviewed in [133]); however, increased incidence of tumors has not been reported in these disorders. This can be exemplified by activating FGFR3-K650E/M mutation, causing thanatophoric dysplasia type II and SADDAN (severe achondroplasia with developmental delay and acanthosis nigricans), respectively [128,129,166,167]. Although this mutation has been detected in aggressive cancers, it failed to induce neoplastic transformation in mice. Additional mutation, involving deletion of the tumor suppressor PTEN (phosphatase and tensin homolog) or activating KRAS mutation were required to induce the FGFR3 cancerogenesis [168,169]. These data suggest that FGFR missense mutations are not likely to initiate the neoplastic transformation, but rather occur later to promote tumor progression and metastasis.
A gene fusion originates from the chromosomal rearrangement involving two genes, and results in a fusion protein capable of neoplastic transformation and oncogene addiction [170,171]. FGFR fusions are relatively rare, accounting for 8% of all FGFR aberrations found in cancer [153,172]. Additional missense mutations are sporadic [172], suggesting that the FGFR fusion protein holds sufficient oncogenic properties. In type I fusions, typically driving the hematological malignancies [173], the FGFR extracellular and transmembrane domains are excluded, and the fusion occurs at the N-terminus of the FGFR kinase domain (Figure 1). In type II fusions that are mostly found in solid tumors [173], the breakpoint usually occurs between exons 17 and 19, affecting only a varying part of the C-terminal region of FGFR [133]. In both types of fusion, the partner typically contains domains that facilitate dimerization such as the coiled-coil domain, the sterile alpha motif, the leucine rich repeat or the leucine zipper, leading to ligand-independent FGFR dimerization and signaling activity. The FGFR fusion protein may also be sequestered to an alternate subcellular location, trough features gained via the fusion partner, which can result in misplaced and deregulated activity. Finally, a substantial part of the fusion partner is typically lost during chromosomal rearrangement, producing haploinsufficiency or gaining novel function that may contribute to neoplastic transformation.
A substantial portion of the FGFR fusion partners are proteins associated with the centrosome functions, including spindle organization and ciliogenesis (8 of 14 recurrent FGFR fusions with at least partially characterized signaling properties; based on PubMed search in April 2021). This led us to speculation that disruption of the centrosome cycle may drive pathogenesis of the FGFR fusion cancers. In the following sections, we review the current knowledge of such oncogenic FGFR fusions, and discuss the possible involvement of both fusion partners in cancerogenesis. For a complete reference, the recurrent and characterized, yet not included fusions comprise FGFR2-CCDC6 [149,174]

FGFR3-TACC3
Gene fusion involving FGFR3 and the transforming acidic coiled-coil containing protein 3 (TACC3) is one of the recurrent gene fusions, found in glioblastoma (29 of 103), nonsmall-cell lung carcinoma (28 of 103), head and neck squamous cell carcinoma (11 of 103), bladder cancer (10 of 103), and other types of cancer (Table 1) [133, 149,153,179,[185][186][187][188][189][190][191][192][193][194][195][196][197][198][199][200]. FGFR3-TACC3 transformed NIH3T3 and Rat1A fibroblasts [179,187,201,202], and the xenografted astrocytes or glioblastoma cells stably expressing FGFR3-TACC3 gave rise to gliomas [187,203]. Mice with hippocampal cells transduced with FGFR3-TACC3 developed invasive, rapidly growing high-grade gliomas [187], proposing FGFR3-TACC3 as an oncogenic driver. The wild-type FGFR comprises the extracellular immunoglobulin-like domain, responsible for ligand binding, the transmembrane region, and the intracellular part that is responsible for binding and activation of the signal transducers including PLCγ (binding site indicated in green). The type II fusions lose a variable part of the C-terminal region of FGFR, frequently involving the PLCγ binding site, and attach a truncated C-terminal part of the fusion partner. In type I fusions, the FGFR extracellular and transmembrane parts are excluded, and the truncated fusion partner joins in just before the FGFR kinase domain. In both types of FGFR fusion, the partner possesses domains that facilitate dimerization-the coiled-coil domain, the sterile alpha motif, the leucine rich repeat or the leucine zipper. The positions of the fusion breakpoints are indicated. The chromosomal rearrangement produces loss of the FGFR3 3 UTR containing miR-99a that normally regulates the FGFR3 levels; this leads to overexpression of FGFR3-TACC3 [203] and abundant transactivation of the FGFR3 residues [201]. Similar to the majority of the type II FGFR fusions, the FGFR3-TACC3 protein lacks the C-terminus of FGFR3 that is necessary for phospholipase C γ (PLCγ) binding (Figure 1), leading to silencing of this signaling branch [179,249]. Conversely, the ERK MAP kinase and STAT (signal transducer and activator of transcription proteins) signaling is increased in FGFR3-TACC3 expressing cells [201,203], and silencing of these pathways was partially successful in targeting the oncogene-driven growth of cell lines and xenografts [36,149,186,187,202,203,[250][251][252].
TACC3 is an important component of the mitotic spindles, ensuring proper attachment of chromosomes to the microtubules [253,254]. During mitosis, FGFR3-TACC3 mislocalizes to the spindle poles while sequestering also the endogenous TACC3 from the mitotic spindle, through interaction of their coiled-coil domains [188, 255,256]. This delays mitotic progression, and induces chromosome segregation defects and aneuploidy that increases by greater than 2.5 fold [187]. Interestingly, targeting TACC3 proved a viable strategy in TACC3-overexpressing cancers, likely by inducing abundant multipolar spindles, which led to mitotic arrest and apoptosis [257][258][259]. Elevated cellular levels of TACC3 were shown to induce loss of primary cilia through Aurora A induction and disruption of the transmembrane protein 67 (TMEM67)-filamin A complex [260,261], and promoted oncogenic transformation and shortened survival of the patients with prostate cancer [262]. Knockdown of TACC3 rescued ciliogenesis, reduced transformation and inhibited xenograft growth [262]. Taken together, FGFR3-TACC3 could lead to neoplastic transformation partly through induction of cilia disassembly and deregulated cell division, which are both druggable targets.

FGFR1-TACC1
The FGFR1 fusion with transforming acidic coiled-coil containing protein 1 (TACC1) was found in various types of tumors arising within the central nervous system (14 of 15;  263], and the xenografted astrocytes stably expressing FGFR1-TACC1 gave rise to gliomas [187]. The biological and oncogenic functions of FGFR1-TACC1 appear similar to those assigned to FGFR3-TACC3 [187]. TACC1 has a coiled-coil domain at the C-terminus, that is preserved in the fusion protein (Figure 1), and that mediates localization to the mitotic spindle [264][265][266]. FGFR1-TACC1 expression increased the rate of errors in chromosomal segregation about five times [187], likely through mislocalization and sequestration of endogenous TACC1, and similar spindle defects were observed in HeLa cells with depleted TACC1 [266]. TACC1 interacts with Aurora A, which appears critical for spindle formation, and the expression levels of the two proteins seem to correlate in cancers [266]. This suggests that TACC1 overexpression caused by FGFR1-TACC1 fusion could participate in neoplastic transformation through deciliation caused by increased Aurora A activity and deregulated cell division, similar to FGFR3-TACC3 cancers.
As a consequence of the chromosomal rearrangement, the FGFR2 3 UTR is truncated which results in upregulation of the FGFR2-BICC1 fusion protein [214]. FGFR2-BICC1 dimerizes likely via the sterile alpha motifs of BICC1 [268], leading to ligandindependent dimerization [149] and activation of the ERK MAP kinase, but not STAT3 or AKT signaling [175, 212,267]. FGFR inhibitors were partially successful in targeting the oncogene-driven growth of cell lines, xenografts and patients' tumors [175,215,269,270]; acquired resistance through gatekeeper FGFR2-V564F mutation was also reported [270]. The FGFR2 V546F -BICC1 cells showed oncogene addiction that was fully inhibited by a synergistic effect of the FGFR and ERK MAP kinase pathway inhibitors [267].
BICC1 is a conserved RNA-binding protein that represses translation of selected mRNAs to control development [271][272][273][274][275]; the domains responsible for RNA binding are, however, partly lost during the chromosomal rearrangement, suggesting that this function is lost with the FGFR2-BICC1 fusion. Deletion of BICC1 leads to classical ciliopathy features, including randomization of the left-right asymmetry, and cystic development in the kidney, liver and pancreas [276][277][278][279][280][281][282][283]. Loss of BICC1 disrupted alignment of motile cilia and establishment of the cilia-driven fluid flow in the mouse embryonic node and Xenopus gastrocoel [279], producing laterality defects. This may be due to disrupted protein synthesis machinery at the centrosome that appears important for the adjacent cilia [284,285]. In humans, mutations in BICC1 were identified in patients with kidney dysplasia, likely caused by ectopic Wingless-related integration site (WNT)/β-catenin signaling [286]. Decreased levels of BICC1, or loss of some of the three RNA-binding domains which are also relevant for the FGFR2-BICC1 fusion, also upregulated WNT/βcatenin signaling [275,279,[287][288][289]. Taken together, the FGFR2-BICC1 fusion is likely to produce a BICC1 haploinsufficiency that leads to disrupted ciliogenesis and cilia-associated signaling, which may contribute to cancerogenesis.

FGFR2-NDC80
A cholangiocarcinoma patient was described with a fusion comprising FGFR2 and NDC80 (or HEC1, highly expressed in cancer 1) [216]. FGFR2-NDC80 was overexpressed in the tumor cells, and activated the ERK MAP kinase, PLCγ, and STAT3 signaling [216]. Considering the PLCγ binding site is lost with the fusion (Figure 1), it is possible that FGFR2-NDC80 activates this pathway through heterodimerization with the endogenous FGFR. The fusion protein retains the kinetochore microtubule binding region of NDC80 [290], sug-gesting possible mislocalization that was, however, not experimentally addressed; within the tumor samples, FGFR2-NDC80 localized predominantly to the cell membrane [216].
NDC80 localizes to the centrosomes and mitotic spindles where it is necessary for assembly and stabilization of the kinetochore microtubules (reviewed in [290]). High NDC80 levels were found in cancers [291][292][293][294], and overexpression of NDC80 in mice led to abnormal spindle formation, hyperactivation of the mitotic checkpoint and initiation of the tumorigenic events [295]. Depletion or inhibition of NDC80 induced mitotic arrest, and suppressed xenograft tumor growth [294,[296][297][298]. Taken together, these data suggest a possible involvement of mitotic defects in the FGFR2-NDC80 cancerogenesis, through ectopic FGFR and NDC80 activity.

FGFR2-OFD1
Fusions involving FGFR2 and the oral-facial-digital type 1 (OFD1) gene were reported in thyroid and endometrial cancer [149,219] (Table 1). FGFR2-OFD1 induced transformation of RK3E cells, that was abolished by FGFR kinase inhibitors [316]. Dimerization of the fusion protein likely occurs through the coiled-coil domains of OFD1 [149], which are preserved in the fusion protein ( Figure 1), leading to transactivation of the FGFR2 kinase domain and activated ERK MAP kinase signaling [316].
OFD1 localizes to centrosome [317] where it is required for centriole maturation and primary ciliogenesis [318,319]. This localization requires the N-terminal part of OFD1 [320] that is, however, lost in the FGFR2-OFD1 fusion. Heterozygous loss-of-function mutations in OFD1 produce the OFD1 syndrome, an X-linked dominant disorder lethal in males that is characterized by systemic ciliopathy features [306,[321][322][323][324]. The Ofd1 +/− female mice reproduced the main patient phenotypes [318,325], suggesting haploinsufficiency in the heterozygous animals. The cilia were severely disrupted or lost, producing defects in laterality and Hh-dependent tissue patterning [318,326]. The zebrafish ofd1 morphants also displayed laterality defects, due to cilia abnormalities in the Kupffer s vesicle, as well as additional ciliopathy features [327]. These data suggest that the decreased levels of endogenous and centrosome-competent OFD1 in the FGFR2-OFD1 cancers may lead to deregulated ciliogenesis and cilia signaling, potentially contributing to neoplastic transformation.
The FOP haploinsufficiency may contribute to FOP-FGFR1 cancerogenesis, as reduced FOP levels were shown to disrupt the centrosome structure and inhibit ciliogenesis [341][342][343], and similar defects were observed in FOP-FGFR1 expressing cells [227,340]. Although the hematopoietic cells do not produce cilia [344,345], the centrosome defects have also been associated with other myeloproliferative neoplasms [340,346], suggesting a common pathogenesis.

CEP110-FGFR1
The fusion of FGFR1 with the centrosomal protein 110 (CEP110) drives expansion of the hematopoietic stem cell population, and causes malignancies that frequently turn into AML [221, (Table 1). When expressed in cells, CEP110-FGFR1 likely dimerizes through the leucine zippers in CEP110 ( Figure 1) which drives constitutive autophosphorylation of the FGFR1 kinase domains [247]. CEP110-FGFR1 induced oncogene addiction in Ba/F3 cells [241,347,348], that could be targeted by tyrosine kinase inhibitors [241,348]. Transplantation of murine bone marrow or human CD34+ cord blood cells transduced with CEP110-FGFR1 produced AML in the recipient mice [347], further supporting the role of CEP110-FGFR1 as an oncogenic driver.
Pluripotent stem cells derived from the AML CEP110-FGFR1 patient showed aberrant hematopoietic differentiation, which was restored by tyrosine kinase inhibitors; a growth inhibition was also achieved with isolated primary AML CEP110-FGFR1 cells [240]. This is in a sharp contrast with the clinical observation, as patients with CEP110-FGFR1 disease do not respond to tyrosine kinase inhibitors and have particularly poor prognosis; allogeneic hematopoietic stem cell transplantation appears the only viable option [238,349]. These data suggest that inhibition of the ectopic FGFR1 kinase activity in CEP110-FGFR1 cancers [241,350] does not bring clinical benefits, and that perhaps additional mechanisms contribute to the disease pathogenesis.
CEP110 is a structural protein of the centrosome [351,352], for which it requires a 170aa region in the C-terminus that is retained in the CEP110-FGFR1 fusion ( Figure 1) [247]. The centrosome localization of the fusion may, therefore, interfere with centrosome maturation, likely due to combination of the steric effects of the fusion and its ectopic kinase activity, which in turn produces centrosomal and spindle abnormalities and drives the oncogenesis [351,353,354].

Conclusions and Perspectives
The FGFR fusion proteins are oncogenic drivers; therefore, patients typically show a good initial response to the targeted therapy using FGFR tyrosine kinase inhibitors [171,186,215,219,269,270,355]. However, secondary gatekeeper mutations occur during therapy [270,356], and inhibition of effectors downstream from the FGFR oncogene has not delivered strong clinical benefit; therefore, alternate approaches are being developed. One such strategy takes advantage of the general overexpression of type II FGFR fusion proteins [268], which makes them a good target for cytotoxic conjugates specifically binding FGFR. For example, FGF2 conjugated with auristatin induced endocytosis of the FGFR1-FGF2/auristatin complexes, which released auristatin and produced a strong cytotoxic effect on cancer cells overexpressing FGFR1 [357]. Similarly, the FGFR-specific antibodies or antibody fragments conjugated to a cytotoxic molecule enter the cells via endocytosis to induce cell death [358,359]. Clinical trials evaluating cytotoxic conjugates in FGFR fusion-driven cancers are yet to emerge.
Another possibility is to specifically target the fusion protein. For example, no therapy protocol is available for FOP-FGFR1-driven cancers, which are very aggressive [221,222,328,331]. FOP-FGFR1 saturates at the centrosome, which appears critical for oncogenic transformation [329,331]. An adeno-associated virus-mediated delivery of interfering RNA, peptide or a coding sequence, specifically targeting the FOP-FGFR1 fusion or its interaction interface with the centrosome, therefore represents an attractive therapeutic possibility [360][361][362].
Finally, the ectopic activity of the FGFR fusion protein, together with decreased levels of the endogenous fusion partner, may contribute to neoplastic transformation through loss of primary cilia and deregulated cell division. Restoration of ciliogenesis and/or cilia function is, therefore, an attractive and so far unappreciated strategy to attenuate tumor growth. NSC12, an orally available analog of the naturally occurring FGF ligand trap pentraxin 3 (PTX3), was developed to target the FGF-driven pathologies [363]. NSC12 rescued ciliogenesis defects in three FGFR-driven cancer cell lines and a xenograft, and inhibited tumor growth [363]. The clinical studies evaluating cilia targeting as a cancer therapy are however yet to emerge.