Oncogenic kinase fusions: an evolving arena with innovative clinical opportunities.

Cancer biology relies on intrinsic and extrinsic deregulated pathways, involving a plethora of intra-cellular and extra-cellular components. Tyrosine kinases are frequently deregulated genes, whose aberrant expression is often caused by major cytogenetic events (e.g. chromosomal translocations). The resulting tyrosine kinase fusions (TKFs) prompt the activation of oncogenic pathways, determining the biological and clinical features of the associated tumors. First reported half a century ago, oncogenic TKFs are now found in a large series of hematologic and solid tumors. The molecular basis of TKFs has been thoroughly investigated and tailored therapies against recurrent TKFs have recently been developed. This review illustrates the biology of oncogenic TKFs and their role in solid as well as hematological malignancies. We also address the therapeutic implications of TKFs and the many open issues concerning their clinical impact.

INTRODUCTION

The discovery of Philadelphia chromosome (Ph) in Chronic Myeloid Leukemia (CML) [1], the subsequent identification of the genes involved in this translocation [2] and the description of its molecular mechanisms [3, 4] have marked a scientific milestone, which dramatically shaped our understanding of kinase-mediated oncogenesis.

Since these findings, many studies have investigated cell transformation via Tyrosine Kinase Fusions (TKFs). A recent appraisal of the literature shows that ~1000 independent studies have been published describing the oncogenic properties of TKFs. Recurrent kinase chimeras are detected in ~2-3% of human tumors and this frequency increases when non-recurrent fusions (4-5%) or cancer-associated abnormal karyotypes (5-7%) are considered [5]. TKFs do not represent the only cancer-associated gene derangement involving Tyrosine Kinases (TKs). During oncogenic transformation, TKs can indeed undergo a number of molecular impairments, including: (i) gene mutations; (ii) gene amplifications; (iii) chromosomal translocations; and (iv) non-fusion rearrangements (e.g. CRLF2 rearrangements in ALL) [6, 7]. All such Tyrosine Kinase (TK)-associated aberrations induce a huge variety of metabolic and signaling derangements, which ultimately lead to neoplastic transformation.

The widespread occurrence of TKFs in human cancers, their ability to synergistically fire multiple signaling pathways and their overall pivotal role in clinical oncology justify the many recent efforts to identify new anti-TKF target therapies. Of note, similar TKF-related pathways have been reported in different tumors; this constitutes the rationale for the use of similar therapies in tumors bearing the same molecular derangements. This approach is well represented by Ph-like leukemias, whose transcriptomes closely resemble those of conventional Ph-positive ALL, even in the absence of BCR-ABL translocations [8, 9]. This model and its corresponding Patient Derived Tumor Xenografts (PDTX) allow the assessment of therapeutic agents targeting common molecular “hubs” [10]. This perspective is in line with the new “basket trial” era, where patients with different tumor types may be grouped together based on their tumors' mutational profile.

In this review, the role of TKFs in both hematologic and solid neoplasms will be discussed. Special attention will be devoted to the many open issues concerning the treatment of TKF-bearing neoplasms.

STRUCTURES OF TKFS AND ROLE OF THE N-TERMINUS DOMAINS

Tyrosine kinases are the most common kinases and catalyze the transfer of a phosphate group from ATP to tyrosine residues. On a structural level, TK display a variety of modules dictating their cellular localization, function stability and response to therapy [11]. The enzymatic activity of these proteins relies on their TK domains (TKD), which are targeted by ATP-competitive and irreversible, covalent inhibitors [12]. Other important regions are the SH2 domains, the extracellular domains and remote allosteric sites, which can also be targeted by neutralizing drugs [13].

TKD are highly conserved modules, frequently composed by a N-terminal ATP-binding region and by a regulatory C-terminal domain. Upon activation, a cascade of events occurs, including ATP consumption, kinase dimerization, conformational changes, trans-autophosphorylation and oligo-complex formation. TK regulation is disrupted in oncogenic TKFs, which display constitutive dimerization due to loss of the inhibitory domains and forced oligodimerization of partner peptides [14]. Both inter- and intra-chromosomal rearrangements can lead to TKFs. Most break-points occur within introns and allow the synthesis of chimeric transcripts carrying the functional kinase domain and its flanking regions (Tables 1 and 2). Translocations involving TK membrane receptors frequently have breakpoints between exons coding for the juxta-membrane region. Though less frequent, the breaks can also split the trans-membrane domain upstream from the GXGXXG-coding exon. If protein products presenting the TK transmembrane domain localize to the cellular membrane, the exclusion of such a domain will localize the TK according to the properties of the translocation partner instead (Figure 1). In both cases, the natural TK ligand-binding site is lost. TKF involving intra-cytoplasmic kinases are also characterized by the loss of regulatory regions.

Table 1

Tyrosine kinases fusion in hematological malignancies

Kinase	Activating Mechanisms	Chromosomal translocation	Entity	Kinase inhibitor	Frequency (%)	Reference
ALK	Fusion to NPM1	t(2;5)(p23;q35)	ALCL, DLBCL	Crizotinib	75-80, N/A	Morris et al. 1994, Adam et al. 2003
	Fusion to ALO17	t(2;17)(p23;q25)	ALCL	Crizotinib	<1	Cools et al. 2002
	Fusion to TGF	t(2;3)(p23;q21)	ALCL	Crizotinib	2	Hernandez et al. 1999
	Fusion to MSN	t(2;X)(p32;q11-12)	ALCL	Crizotinib	<1	Tort et al. 2001
	Fusion to TPM3	t(1;2)(q25;p23)	ALCL	Crizotinib	12-18	Lamant et al. 1999
	Fusion to TPM4	t(2;19)(p23;p13)	ALCL	Crizotinib	<1	Meech et al. 2001
	Fusion to ATIC	inv(2)(p23;q35)	ALCL	Crizotinib	2	Ma et al. 2000
	Fusion to MYH9	t(2;22)(p23;q11.2)	ALCL	Crizotinib	<1	Lamant et al. 2003
	Fusion to TRAF1	t(2;9)(p23;q33)	ALCL	Crizotinib	<1	Feldman AL et al. 2013
	Fusion to CLTC1	t(2;17)(p23;q23)	ALCL, DLBCL	Crizotinib	2, N/A	Touriol et al. 2000
	Fusion to SQSTM1	t(2;5)(p23.1;q35.3)	DLBCL	Crizotinib	N/A	Takeuchi et al. 2010
	Fusion to SEC31A	t(2;4)(p24;q21)	DLBCL	Crizotinib	N/A	Bedwell et al. 2007
	Fusion to RANBP2	inv(2)(p23;q13)	AML	Crizotinib	<1	Maesako et al. 2014
ABL	Fusion to BCR	t(9;22)(q34;q11)	CML, B-ALL	Imatinib	85-90, <30	Klein et al. 1982
	Fusion to TEL	t(9;12)(q34;p13)	AML	Dasatinib	<1	Golub et al. 1996
	Fusion to NUP214	t(9;9)(q34.1;q34.3)	T-ALL, Ph-like ALL	Nitolinib	5	Graux et al. 2004, Roberts et al 2014
	Fusion to EML1	t(9;14)(q34;q32)	T-ALL	Nitolinib	<1	De Keersmaecker et al. 2005
	Fusion to ZMIZ1	t(9;10)(q34;q22)	B-ALL	N/A	<1	Soler et al. 2008
	Fusion to RCSD1	t(1;9)(q24;q34)	B-ALL	Dasatinib	<1	Mustjoki et al. 2009
	Fusion to FOXP1	t(3;9)(p12;q34)	B-ALL	N/A	<1	Ernst T et al. 2011
	Fusion to SNX2	t(5;9)(q23;q34)	B-ALL	Imatinib	<1	Ernst T et al. 2011, Masuzawa et al. 2014
	Fusion to SEPT9	t(9;17)(q34;q25)	T-PLL	N/A	<1	Suzuki et al. 2014
	Fusion to multiple partners	t(9;12)(q34;p13)	Ph-like ALL	Dasatinib	<1	Roberts et al. 2012, Roberts et al. 2014
ARG	Fusion to TEL	t(1;12)(q25;p13)	AML, aCML	Imatinib	<1	Golub et al. 1995
PDGFRa	Fusion to FIP1L1	t(4;12)(q23;p12)	HES	Imatinib	12	Cools et al. 2003
	Fusion to BCR	t(4;22)(q12;q11.2)	CEL, T-ALL, CML-like MPN	Imatinib	<1	Baxter et al. 2002, Yigit et al. 2015, Cluzeau et al. 2015
	Fusion to TNKS2	t(4;10)(q12;q23.3)	MPN w/eosinophilia	Imatinib	<1	Chalmer et al. 2014
	Fusion to STRN	t(2;4)(p22;q12)	MPN w/eosinophilia	Imatinib	<1	Curtis et al. 2007
	Fusion to ETV6	t(4;12)(q23;p12)	MPN w/eosinophilia	Imatinib	<1	Curtis et al. 2007
	Fusion to KIF5B	t(4;10)(q12;p11)	MPN w/eosinophilia	Imatinib	<1	Score et al. 2006
	Fusion to CDK5RAP2	ins(9;4)(q33;q12q25)	CEL	Imatinib	<1	Walz et al. 2006
PDGFRb	Fusion to TEL	t(5;12)(q33;p13)	CMML	Imatinib	4	Golub et al. 1994
	Fusion to HIP1	t(5;7)(q33;q11)	CMML	Imatinib	4	Ross et al. 1998
	Fusion to Rabaptin5	t(5;17)(q33;p13)	CMML	Imatinib	<1	Magnusson et al. 2001
	Fusion to H4(D10S170)	t(5;10)(q33;q11-q21)	aCML	Imatinib	<1	Kulkarni et al. 2000
	Fusion to CEV14	t(5;14)(q33;q32)	AML	Imatinib	<1	Abe et al. 1997
	Fusion to Myomegalin	t(1;5)(q23;q33)	Eosinophilia	Imatinib	<1	Wilkinson et al. 2003
	Fusion to ATF71P	t(5;12)(q23;p13)	Ph-like ALL	Imatinib	<1	Kobyashi et al. 2015
	Fusion to EBF1	t(5;5)(q33.1;q33.3)	Ph-like ALL	Dasatinib	<1	Roberts et al. 2012, Roberts et al. 2014
FGFR1	Fusion to multiple partners	t(2;8)(q12;p11)	EMS, AML	None	<1	Etienne et al. 2007
	Fusion to FOP	t(6; 8)(q27;p11)	MPN	None	<1	Lee et al. 2014
	Fusion to SQSTM1	t(5;8)(q35;p11)	AML		<1	Nakamura Y et al. 2014
FGFR3	Fusion to TEL	t(4;12)(p16;p13)	PTCL	Fiin23, NVP-BGJ398	<1	Maeda et al. 2005
	Fusion to IGH	t(4;14) (p16; q32)	CLL		<1	Geller et al. 2014
	Fusion to TIF1	t(7;8)(q34;p11)	MDS, CLL, AML	Fiin23, NVP-BGJ398	<1	Maeda et al. 2005
JAK2	Fusion to TEL	t(9;12)(p24;p13)	ALL, CML-like	Ruxolitinib	<5	Lacronique et al. 1997
	Fusion to OFD1	t(X;9)(p22;p24)	ALL	Jak2 inhibitors	<1	Yano et al. 2015
	Fusion to SPAG9	t(9;17)(p24;q21)	ALL	Jak2 inhibitors	<1	Kavamura M et al 2015
	Fusion to PAX5	t(9;9)(p13;p24)	ALL	Jak2 inhibitors	<1	Nebral K et al. 2009
	Fusion to BCR	t(9;22)(p24;q11.2)	aCML	Ruxolitinib	<5	Griesinger et al. 2005
	Fusion to multiple partners	t(9;12)(p24;p13)	Ph-like ALL	Jak2 inhibitors	<1	Roberts et al. 2012, Roberts et al. 2014
NTRK3	Fusion to TEL	t(12;15)(p13;q25)	AML	None	<1	Knezevich et al. 1998
SYK	Fusion to TEL	t(9;12)(q22;p12)	MDS	Imatinib	<1	Kanie et al. 2004
	Fusion to ITK	t(5;9)(q33;q22)	PTCL-NOS, AITL	None	17, <1	Streubel B et al 2006, Attygale et al 2013
TRKC	Fusion to TEL	t(12;15)(p13;q25)	AML and fibrosarcome	None	<1	Dobus et al. 2001
FLT3	Fusion to ETV6	t(12;13)(p13;q12)	HES	Sunitinib, Midostaurin, Lestaurtinib	<1	Vu et al. 2006
ETV6	Fusion to TEL	t(4;12)(p16;p13)	PTCL-NOS	Ruxolitinib	<1	Yagasaki et al. 2001
CSF1R	Fusion to SSBP2	t(5;5)(q14;q33)	Ph-like ALL	Dasatinib	<1	Roberts et al. 2014
CRLF2	Fusion to IGH	t(X;14)(p22;q32)/t(Y;14)(p11;q32)	Ph-like ALL	Jak2 inhibitors	<5	Mullighan et al. 2009, Roberts et al. 2014
ROS1	Fusion to NFKB2	t(6;10)(q22;q24)	ALCL	Ros1 inhibitors	<1	Crescenzo et al. 2015
	Fusion to NCOR2	t(6;12)(q22;q24)	ALCL	Ros1 inhibitors	<1	Crescenzo et al. 2015
TYK2	Fusion to NFKB2	t(19;10)(p13;q24)	ALCL	Tyk2 inhibitors	<1	Crescenzo et al. 2015
	Fusion to NPM1	t(19;5)(p13;q35)	LPDs	Tyk2 inhibitors	<1	Velusamy et al. 2014
LYN	Fusion to NCOR	t(8;17)(q13;p11)	ALL	NA	<1	Yano et al. 2015

Abbreviation: ALCL: anaplastic large cell lymphoma, AML: acute myeloid leukemia, B-ALL: B-cell acute lymphoblastic leukemia; T-ALL: T-cell acute lymphoblastic leukemia, Ph-like ALL: Philadelphia Chromosome like acute lymphoblastic leukemia, CEL: chronic eosinophilic leukemia, CML: chronic myeloid leukaemia; aCML: atypical chronic myeloid leukemia, CMML: chronic myelomonocytic leukemia, DLBCL: diffuse large B-cell lymphoma, EMS: 8p11 myeloproliferative syndrome, HES: hyper eosinophilic syndrome, LPDs: lymphoproliferative disorders, MDS: myelodysplastic syndrome, MPN: myeloproliferative neoplasm, PTCL-NOS: peripheral T-cell lymphoma not otherwise specified

Table 2

Tyrosine kinases fusions human in solid tumors

Kinase	Activating Mechanisms	Chromosomal translocation	Entity	Kinase inhibitor	Frequency (%)	Reference
ALK	Fusion to ATIC	inv(2)(p23;q35)	IMT	Crizotinib	< 5	Debiec-Rychter et al. 2003
	Fusion to CARS	t(2;11)(p23;p15)	IMT	Crizotinib	<5	Cools et al. 2002
	Fusion to CLTC	t(2;17)(p23;q23)	IMT	Crizotinib	<5	Bridge et al. 2001
	Fusion to EML4	inv(2)(p21;p23)	NSCLC	Crizotinib, Ceritinib, Alecitinib	2-5	Soda et al. 2007
			BC, CRC	Crizotinib	<5	Lin et al. 2009
	Fusion to FN1	t(2;11)(q31;p15)	Soft tissue sarcoma	Crizotinib	2-4	Ren et al. 2012
	Fusion to KIF5B	t(2;10)(p23;p11)	NSCLC	Crizotinib	<1	Takeuchi et al. 2009
	Fusion to KLC1	t(2;14)(p23;q32)	NSCLC	Crizotinib	<5	Jung et al. 2012
	Fusion to RANBP2	t(2;2)(p23;q13)	IMT	Crizotinib	<5	Ma et al. 2003
	Fusion to SEC31L1	t(2;4)(p23;q21)	IMT	Crizotinib	<5	Panagopoulos et al. 2006
	Fusion to VCL	t(2;10)(p23;q22)	RCC	Crizotinib	<3	Debelenko et al. 2011
	Fusion to SEC31A	t(2;4)(p23;q21)	NSCLC	Crizotinib	<1	Kim et al. 2015
	Fusion to STRN	t(2;2)(p23;p22)	Thyroid cancer	Crizotinib	<1	Pérot et al. 2014, Kelly et al. 2013
			NSCLC	Crizotinib	<1	Majewski et al. 2013
	Fusion to GTF2IRD1	t(2;7)(p23;q11.23)	Thyroid cancer	Crizotinib	<1	Stransky et al. 2015
	Fusion to TFG	t(2;3)(p23;q21)	NSCLC	Crizotinib	2	Rikova et al. 2007
	Fusion to TPM1	t(2;15)(p23;q22.2)	Bladder cancer	Crizotinib	<1	Stransky et al. 2015
	Fusion to TPM3	t(1;2)(q21;p23)	IMT	Crizotinib	50	Lawrence et al. 2000
	Fusion to TPM4	t(2;19)(p23;p13)	IMT	Crizotinib	<5	Lawrence et al. 2000
	Fusion to PTPN3	t(2;9)(p23;q31.3)	NSCLC	Crizotinib	<1	Jung et al. 2012
	Fusion to A2M	t(2;12)(p23;p13)	FLIT	Crizotinib	<1	Onoda et al. 2014
	Fusion to TPR	t(2;1)(p23;q31.1)	NSCLC	Crizotinib	<1	Choi et al. 2014
	Fusion to HIP1	t(2;7)(p23;q11.23)	NSCLC	Crizotinib	<1	Hong et al. 2014
	Fusion to SQSTM1	t(2;5)(p23;q35)	NSCLC	Crizotinib	<1	Iyevleva et al. 2015
	Fusion to DCTN1	t(2;2)(p23;p13)	NSCLC	Crizotinib	<1	Iyevleva et al. 2015
	Fusion to SMEK2	t(2;2)(p23;p16.1)	CRC	Crizotinib	<1	Stransky et al. 2015
	Fusion to CAD	inv(2)(p22-21p23)	CRC	Entrectinib	<1	Lee et al. 2015, Amatu et al. 2015
ROS1	Fusion to CD74	t(5;6)(q32;q22)	NSCLC	Crizotinib	<2	Bergethon et al. 2012
	Fusion to EZR	inv(6)(q22q25.3)	NSCLC	Crizotinib	<2	Arai et al. 2013
	Fusion to GOPC	del(6)(q22q22.3)	NSCLC	Crizotinib	<2	Rimkunas et al. 2012, Suehara et al. 2012
			CCA	Crizotinib	<1	Gu et al. 2011
			Ovarian Cancer	Crizotinib	<1	Birch et al. 2011
	Fusion to LRIG3	t(6;12)(q22;q14.1)	NSCLC	Crizotinib	<2	Takeuchi et al. 2012
	Fusion to SDC4	t(6;20)(q22;q12)	NSCLC	Crizotinib	<2	Davies et al. 2012, Takeuchi et al. 2012
	Fusion to SLC34A2	t(4;6)(q15.2;q22)	NSCLC	Crizotinib	<2	Davies et al. 2012
			Gastric cancer	Crizotinib	<1	Lee et al. 2013
	Fusion to TPM3	t(1;6)(q21.2;q22)	NSCLC	Crizotinib	<2	Takeuchi et al. 2012
	Fusion to TFG	t(6;3)(q22.1;q12.2)	IMT	Crizotinib	<1	Yamamoto et al. 2015
RET	Fusion to CCDC6	inv10(q11;q21)	NSCLC	Cabozantinib, Vandetanib	<2	Wang et al. 2012
			Thyroid cancer	Cabozantinib, Vandetanib	<2	Celestino et al. 2012
	Fusion to KIF5B	inv(10)(p11;q11)	NSCLC	Cabozantinib, Vandetanib	<2	Ju et al. 2012
	Fusion to NCOA4	inv(10)(q11;q11)	Thyroid cancer	Cabozantinib, Vandetanib	<2	Rui et al. 2012
	Fusion to PRKAR1A	t(10;17)(q11.2;q23)	Thyroid cancer	Cabozantinib, Vandetanib	<2	Rui et al. 2012
	Fusion to ACBD5	inv(10)(p12.1;q11.2)	Thyroid cancer	Cabozantinib, Vandetanib	<1	Hamatani et al. 2014
BRAF	Fusion to KIAA1549	t(7;7)(q34;q34)	Brain tumors	BRAF/MEK inhibitors	<1	Tian et al. 2011
	Fusion to FAM131B	t(7;7)(q34;q34)	Brain tumors	BRAF/MEK inhibitors	<1	Cin et al. 2011
	Fusion to CEP89	t(7;19)(q34;q13)	Melanoma	BRAF/MEK inhibitors	< 5	Wiesner et al. 2014
	Fusion to LSM14A	t(7;19)(q34;q13)	Melanoma	BRAF/MEK inhibitors	< 5	Wiesner et al. 2014
FGFR1	Fusion to TACC1	t(8;8)(p11.23;p11.22)	GBM	FGFR inhibitor	<3	Singh et al. 2012
	Fusion to BAG4	t(8;8)(p11.23;p11.23)	NSCLC	FGFR inhibitor	<1	Rui et al. 2014
FGFR2	Fusion to BICC1	t(10;10)(q26;q21.1)	CCA	FGFR inhibitor	<1	Yi-Mi et al. 2013
	Fusion to KIAA1967	t(10;8)(q26;p21.3)	NSCLC	FGFR inhibitor	<1	Yi-Mi et al. 2013
	Fusion to PPHLN1	t(10;12)(q26;q12)	CCA	FGFR inhibitor	45	Sia et al. 2015
FGFR3	Fusion to TACC3	del4(p16;p16)	GBM	FGFR inhibitor	<3	Singh et al. 2012, Bao et al. 2014
			Bladder cancer	FGFR inhibitor	<2	Williams et al. 2013
			NSCLC	FGFR inhibitor	<2	Rui et al. 2014
			ESCC	FGFR inhibitor	<1	Yuan et al. 2014
			NPC	FGFR inhibitor	<3	Yuan et al. 2014
			Cervical cancer	FGFR inhibitor	<1	Carneiro et al. 2015
NTRK1	Fusion to TPM3	t(1;3)(q21;q11)	Thyroid cancer	TRKA inhibitor	7-8	Beimfohr et al. 1999
			CRC	TRKA inhibitor	<1	Creancier et al. 2015
			HGG	TRKA inhibitor	<1	Wu et al. 2014
	Fusion to TPR	inv1(q23;q21)	Thyroid cancer	TRKA inhibitor	<1	Greco et al. 1999
			CRC	TRKA inhibitor	<1	Creancier et al. 2015
	Fusion to MPRIP	t(1;17)(q21;p11)	NSCLC	TRKA inhibitor	<5	Vaishanvi et al. 2013
	Fusion to CD74	t(1;5)(q21;q32)	NSCLC	TRKA inhibitor	<5	Vaishanvi et al. 2013
	Fusion to RABGAP1L	t(1;1)(q21;q25.1)	CCA	TRKA inhibitor	<1	Ross et al. 2014
	Fusion to SQSTM1	t(1;5)(q21;q35)	NSCLC	Entrectinib	<1	Farago et al. 2015
	Fusion to LMNA	t(1;1)(q21;q22)	Soft tissue sarcoma	LOXO-101	<1	Doebele et al. 2015
			CRC	Entrectinib	<1	Sartore-Bianchi et al. 2015
NTRK2	Fusion to VCL	t(9;10)(q22.1;q22)	HGG	TRKA inhibitor	<1	Wu et al. 2014
	Fusion to AGBL4	t(9;1)(q22.1;p33)	HGG	TRKA inhibitor	<1	Wu et al. 2014
NTRK3	Fusion to ETV6	t(12;15)(p13;q25)	Thyroid cancer	TRKA inhibitor	2-14	Ricarte-Filho et al. 2013
			CFS	TRKA inhibitor	<1	Knezevih et al. 1998
			IMT	TRKA inhibitor	<1	Yamamoto et al. 2015
			GIST	TRKA inhibitor	<1	Brenca et al. 2015
			MASC	TRKA inhibitor	<1	Skalovà et al. 2015
			HGG	TRKA inhibitor	<1	Wu et al. 2014
	Fusion to BTBD1	t(15;15)(p24;q25)	HGG	TRKA inhibitor	<1	Wu et al. 2014
AKT2	Fusion to BCAM	t(19;19)(q13.2;q13.3)	HGSC	AKT2 inhibitor	10	Kannan et al. 2015
PRKACA	Fusion to DNAJB1	t(19;19)(p13.1;p13.2)	FL-HCC	PKA inhibitors	100	Honeyman et al. 2014
PRKD1	Fusion to ARID1A	t(14;1)(q12;p36)	Salivary gland tumor	PRKD1 inhibitor	<3	Weinreb et al. 2014
MET	Fusion to PTPRZ1	t(7;7)(q31.2;q31.3)	GBM	MET inhibitor	15	Bao et al. 2014
PIK3CA	Fusion to TBL1XR1	t(3;3)(q26.3;q26.32)	BC	PIK3CA inhibitor	<1	Stransky et al. 2015

Abbreviation: BC: breast cancer, CCA: cholangiocarcinoma, CFS: congenital fibrosarcoma, CRC: colon-rectal cancer, ESCC: esophageal squamous cell carcinoma, FL-HCC: fibrolamellar hepatocellular carcinoma, FLIT: fetal lung interstitial tumor, GBM: glioblastoma, GIST: gastrointestinal stromal tumor, HGG: high grade glioma, HGSC: high-grade serous ovarian cancer, IMT: imflammatory myofibroblatic tumor, MASC: mammary analogue secretory carcinoma of salivary glands, NPC: nasopharyngeal carcinoma, NSCLC: non-small cell lung cancer

Figure 1

Structure and Signaling Transduction Motifs of Tyrosine Kinase Fusions

The constitutive activation of Tyrosine Kinase Fusion oncoproteins are achieved through multiple mechanisms taking advantage of direct or indirect oligodimerization. Seldom no dimerization are required. Fusion partners can also engage per se oncogenic signaling pathways, directly or indirectly modulating Transcription Factors (i.e. NFkB) and their corresponding genes. Kinase activation induces multiple canonical pathways (PI3K/AKT, JAK/STAT, PLCγ/PKC and RAS/ERK), which regulate genes controlling transcription and providing pro-tumorigenic signals. Compensatory pathways and regulatory modalities may act in place (i.e. miRNA regulation).

Gene fusions typically replace the TK promoter; therefore TKF expression becomes ectopically regulated by the promoter of the partner gene. Partner genes contribute to the oncogenic potential in a number of ways. In most instances, the partner N-terminus region provides dimerization domains, which recruit molecular adaptors and lead to the constitutive trans-phosphorylation and activation of the kinase. TKF partners can also directly activate oncogenic pathways, as observed in the TFG-NTRK1 and TRAF1-ALK fusions, independently activating the NF-kB pathway [15]. We and others have recently demonstrated that a small subset of systemic (<5%) and cutaneous (~20%) Anaplastic Large Cell Lymphomas (ALCL) bears ROS1 or TYK2 fusions. In such cases, the TKF N-terminus region comprises a transcription factor (NFkB) or a transcriptional repressor (NCOR2) fused to ROS1 and TYK2, respectively [16]. These TKFs clearly illustrate how the chimera may also influence the transcriptional activity of the partner genes, perturbing transcription promotion or repression. This, in turn, may lead to a widespread derangement of several metabolic and signaling pathways, which ultimately concur to neoplastic transformation [15, 16]. The N-terminus can also contribute to the neoplastic phenotype by preventing protein degradation through the recruitment of heat shock proteins (Hsp70, Hsp90) [17]. This may constitute a valid therapeutic target, as observed with Hsp90 inhibitors for the treatment of NPM-ALK ALCL [16, 17] and FLT3-positive leukemias [18]. Fusion partners can further contribute to neoplastic transformation by directing TKF localization. In T-ALL, for example, NUP214 of the NUP214-ABL orchestrates TKF nuclear pore localization, which is required for neoplastic transformation [19]. Similarly, the ectopic localization of NPM-ALK and FOP-FGFR1 contributes to oncogenesis by impacting centrosome deregulation and chromosomal instability. Lastly, fusion partners can act as dominant negative mutants. By decreasing the access to wild-type TPM3, TPM3-ALK fusions induce changes in the cytoskeleton organization and confer a higher metastatic capacity [20].

TKF IN HUMAN CANCERS

TKFs are found in many human cancers, although at very different frequencies. Among recurrent translocations, hematologic cancers were first shown to carry rearrangements characterized by TKF expression. Subsequent studies demonstrated that RET [21], TRK [22] or NTRK1 [23] were frequently translocated in a fraction of thyroid carcinomas, as well as other epithelial and soft tissue neoplasms [24, 25]. This list has grown steadily to >300 oncogenic TKF [5] (http://cgap.nci.nih.gov/Chromosomes/Mitelman) (Tables 1 and 2).

Are TKF drivers or passengers in human oncogenesis?

There is strong evidence that both TK and TKFs play an oncogenic role in human cancers [26-28]. Nevertheless, TKF-driven transformation requires multiple genomic defects and the contribution of the host microenvironment to drive oncogenesis. This paradigm is well represented by Chronic Eosinophilic Leukemia (CEL), where the sustained expression of FIP1L1-PDGFRA is necessary (but insufficient) to sustain eosinophils proliferation [29]. In humans, CEL also requires the overexpression of IL-5 through a single nucleotide polymorphism of IL5RA. The same holds true for systemic mastocytosis (SM), in which the constitutive activation of c-KIT alone is insufficient to induce the neoplastic phenotype and requires the stimulation of stem cell factor (SCF) as well [30].

This view is somehow challenged by seminal studies on other myeloproliferative neoplasms (MPNs), like CML, which may in fact represent monogenic disorders. The implantation of BCR-ABL (p210) retroviral-transduced marrow precursors into recipient mice can indeed cause rapidly progressing MPNs, which closely resemble human CML [31]. Of note, this scenario is hardly achievable in mouse models, where the TKF is ectopically expressed by means of more conventional molecular strategies [32, 33]. Moreover, Foley et al. have recently shown that the expression of BCR-ABL alone from the knock-in allele is unable to induce leukemia [34], suggesting that this TKF is per se insufficient to induce a disease fully recapitulating human CML. This latter hypothesis is consistent with the observation of BCR-ABL-positive hematopoietic cells also in healthy individuals [35], but is challenged by the observation that a fraction of CML patients who have achieved a sustained complete molecular remission (CMR) on imatinib can remain in CMR after drug withdrawal [36]. This latter observation raises the possibility that BCR-ABL signaling may not be the only driving force in some cases of CML. Of note, the concomitant loss of p14 (ARF) [37-39], and the critical role of Bcl-6 [40], miR-29 [41], and Raf1 [42] in CML suggest that the co-occurrence of selective defects and/or the full activation of selective pathways are required to achieve the complete neoplastic phenotype. This is line with the recent observation that IKF1-mutated Ph-positive high-risk ALL, in which IKF1 and Arf alterations synergistically promote the development of an aggressive lymphoid leukemia, can be effectively treated with retinoid receptor agonists which potentiate the activity of dasatinib in mouse and human BCR-ABL1 ALL [43].

The evidence that normal T-lymphocytes can undergo neoplastic transformation upon ectopic expression of fusion proteins alone (e.g. NPM-ALK) may further challenge the aforementioned paradigm [44]. Further studies are however needed to clarify this issue, as TKF transcripts have also been documented in healthy T-cells and in normal hematopoietic cells [35, 45]. Taken together, these observations indicate that TKFs alone are probably unable to drive T-cell neoplastic transformation completely and require secondary chromosomal abnormalities/TK activating mutations [46].

TKF expression levels may also play a role in tumor development. This is the case of ALK-translocated tumors: high NPM-ALK expression is indeed seen in ALCL, while lower EML4-ALK levels characterize NSCLC. These differences are mainly due to different transcriptional activities of the partner genes. Lastly, the regulation of TKF expression may depend on miRNAs/translational regulation and TKF degradation. Overall, these data provide empirical support to the concept of global neutrality, or near-neutrality (i.e. very weak selection of the mutant clone in the early phases of neoplastic transformation) for TKFs [47].

Are TKFs required for the maintenance of a neoplastic phenotype?

It is generally believed that TKFs contribute to the maintenance of the neoplastic phenotype. This is well exemplified by the therapeutic success of TK inhibitors (TKi) for the treatment of chronic/indolent hematologic disorders (CML and CEL) [48, 49]. The oncogenic contribution of TKF to tumor maintenance can nevertheless vary according to several variables, including: (i) the properties of each fusion; (ii) the cellular context and the tumor microenvironment; (iii) the disease stage and evolution phase [28, 49]; (iv) the degree of oncogenic addiction to the TKF (e.g. ROS1-positive NSCLCs show better response to Crizotinib than ALK-positive NSLCLs) [50]; (v) the presence of concomitant genomic aberrations [30, 51]. In particular, the occurrence of multiple genetic defects may influence the response to therapy. This is the case of genetically complex disorders (e.g. ALL or end-stage solid cancers), which show a very aggressive clinical behavior and only partial responses to TKi [52]. The same holds true for some cases of ALK+ ALCL. The majority of ALK+ ALCL patients have a fairly good outcome with conventional chemotherapy treatments. However, rapid and even leukemic progression can occur. These are frequently associated with disrupting events, including TP53 and Blimp1 deletions [53] and c-myc translocation/amplifications. The acquisition of secondary events may indeed be linked to a more aggressive phenotype and can damper the addiction to ALK signaling [15]. This mechanism has also been reported in CML in blast crisis [54] and in ALK-positive NSLCLs resistant to Crizotinib. In the first case, BCR-ABL induces an IGF1 autocrine loop via Hck and Stat5b, in the latter case, the IGF-1R signaling synergizes with (and overcome) ALK by engaging the adaptor protein IRS-1 [51]. Moreover, restoration of p16 (INKa) or p14 (ARF) into Ph-positive leukemic cells (both CML in blast crisis and ALL) is able to determine cell growth arrest and/or apoptosis suggesting a pathogenic role for these deletions and their co-operating role with BCR-ABL [39].

TKFs in hematopoietic disorders

Among hematologic malignancies, distinct subsets of acute leukemias, MPNs and non-Hodgkin lymphomas harbor specific TKFs (Table 1).

The most frequent TKFs involve ABL-1 and PDFGRa/b, but translocations of ARG, JAK2, NTRK3 and SYK have been reported, as well. Not all partner genes provide oligodimerization domains, as in the case of the FIP1L1-PDGFRa fusion, which leads to the activation of the kinase moiety by disruption of its juxtamembrane regulatory motif [55]. Alternatively, the partners do not interact directly, but engage intermediate proteins leading to an indirect dimerization (e.g. CTCCL-ALK, NUP214-ABL, FOP-FGFR1).

The t(9;22)(q34;q11) translocation juxtaposes the BCR and ABL1 genes in virtually all cases of CML and in a subset of adult/pediatric ALL. The latter may represent a de novo disease or, as in many adults, the blastic transformation of CML. Of note, adult and (more frequently) pediatric ALL carry other TKFs, which are associated with a Ph-like phenotype [9]. The oncogenic properties of such fusions have been demonstrated in a large variety of in vitro and in vivo systems [9, 56, 57]. The second most common TKFs involve PDGFRα, PDGFRβ and FGFR1. Such TKFs characterize a group of hematological disorders, sharing similar clinico-pathological features, such as peripheral blood eosinophilia, clinical presentation as MPN and/or lymphoid tumors, and, at least for PDGFRα and PDGFRβ fusions, good response to Imatinib. For this reason, the 2008 WHO Classification of hematopietic and lymphoid tumors has introduced a separate diagnostic category for such disorders, referred to as “Myeloid and Lymphoid Neoplasms with Eosinophilia and Abnormalities of PDGFRα, PDGFRβ or FGFR1” [58].

Among mature T-cell neoplasms, subsets of Peripheral T-cell Lymphomas, not otherwise specified (PTCL, NOS) and Angioimmunoblastic T-cell Lymphomas (AITL) display SYK translocations and ITK-SYK fusion proteins [59-61]. ALK chimeras have been reported in a subset of ALCL (ALK-positive ALCL) [59], which are again recognized as a specific entity by the 2008 WHO Classification. Of note, the aberrant expression of ITK-SYK and NPM-ALK fusions in mice can lead to T-cell transformation [61, 62]. In particular, ITK-SYK mediates lymphomagenesis through the N-terminal phosphatidylinositol 3,4,5-trisphosphate-binding pleckstrin homology (PH) domain of ITK and SLP-76. ALK fusions, on the other hand, require the kinase domain and its flanking regions to activate multiple signaling pathways [59, 63]. Novel translocations involving TYK2 and ROS-1 in CD30-positive cutaneous lymphoproliferative disorders [64] and systemic ALK-negative ALCL [16] have been previously described. Interestingly, in addition to NPM-1, NFkB2 and NCOR2 are partners of these new TKFs and contribute to the neoplastic phenotype [16]. Finally, precursor T-cell neoplasms (i.e. T-cell acute lymphoblastic lymphoma [T-ALL]) may also harbor TKFs, which usually involve the ABL gene. Some of such translocations are also observed in B-ALL (e.g. ABL-NUP214), while others seem to be specific of T-ALL (ABL-EML1) [65] (Table 1). Interestingly, several lymphoproliferative disorders also display TK oncogenic mutations, further demonstrating the tumorigenic role of deregulated kinase signaling in hematopoietic disorders [66, 67].

TKFs are less frequent in mature non-Hodgkin B-cell neoplasms, with the notable exception of ALK-positive diffuse large B-cell lymphoma, in which the TKFs bear distinct oncogenic and clinico-pathological features [68] (Table 1).

TKFs in solid tumors

In solid tumors, aberrant TK activity results from diverse mechanisms, including: (i) point mutations; (ii) gene amplifications; (iii) gene overexpression; and (iv) chromosomal translocations [51, 69, 70]. Abnormal TK activity can also result from de novo transcription initiation (ATI) sites, which lead to the synthesis of truncated intracellular TKDs (e.g. ALKATI in melanoma or ERBB4ATI in ALCL) [71, 72].

In recent years, a wide array of recurrent translocations has been reported in solid tumors (Table 2) [14, 28]. In a systematic survey of nearly 7000 samples from the Cancer Genome Atlas, Stransky et al. have depicted an extremely variegated translocation landscape in solid tumors [5], which can display heterogeneous chromosomal translocations, with specific malignancies (such as sarcomas) being characterized by a high percentage of non-recurrent translocations and/or infrequent but recurrent gene fusions.

To date, fusions involving 14 different TKs have been described in solid tumors (Table 1). ALK is translocated in several tumors, including lung, colorectal, breast, renal, thyroid carcinomas and soft tissue tumors.

To date, fusions involving 14 different TKs have been described in solid tumors (Table 1). ALK is translocated in several tumors, including lung, colorectal, breast, renal, thyroid carcinomas and soft tissue tumors. Fusion partners include EML4, TFG, KIF5B, KLC1 and STRN [28]. Rare translocations involving SMEK2 (rectal adenocarcinoma) or GTF2IRD1 (thyroid carcinoma) have also been reported.

Next generation sequencing (NGS) analyses have highlighted new patterns of translocations, defining specific subgroups of common solid tumors (BRAF in melanoma; FGFR1, FGFR2, FGFR3 in lung cancer; NTRK in thyroid cancer), and in clinically unmet patients (BRAF and FGFR1 translocations in pediatric low-grade glioma; NTRK1 and NTRK2 in high-grade gliomas, and cholangiocarcinoma, Table 1). Lastly, MET is involved in chromosomal translocations of secondary glioblastomas and papillary renal carcinomas and PI3KCA translocations are observed in breast and prostate carcinomas.

The transforming activity of TKF in solid cancers has been extensively characterized through pre-clinical models [73], proving that classical MEK-ERK and PI3K-AKT pathways are critical players in sustaining the transformed phenotype and in mediating resistance to targeted therapies [74]. TKF dependency thus provides a molecular explanation for the “oncogene addiction” of different TKF-driven tumors [75].

Mechanistically, multiple causes of chromosomal translocations have been predicted: (i) exposure to toxic agents (probably the major initiating event); (ii) exposure to ionizing radiation (e.g. in papillary thyroid carcinoma); (iii) oxidative stress and chemotherapeutic agents (e.g. topoisomerase inhibitors).

DISCOVERY AND DETECTION OF TK FUSIONS

The oncogenic role of TKFs has prompted huge efforts to improve their detection in cancer patients to better guide patient-directed treatment strategies [76].

Affinity purification coupled with mass spectrometry (AP-MS) is a powerful tool to characterize aberrant gene products and protein-protein interactions (PPIs). AP-MS has been successfully used to discover novel TKFs in NSCLC (ALK and ROS1) [77] and rare TKFs in other tumor histotypes [78]. AP-MS could also define activating mutations linked to TKi-resistant phenotypes [79]. Furthermore, proteomic patterns distinguish normal from pathological tissues and correctly classify primary/metastatic lesions, allowing tailored treatments and a better stratification of NSCLC [80].

The introduction of NGS-based approaches has rapidly enriched the overall knowledge of TKF, giving a better representation of their distribution in human cancers. In hematologic malignancies, NGS studies of Ph-like ALL have detected TKFs in >90% of patients [8]. Technically, both RNAseq and Whole Exome Sequence (WES) made these achievements possible. In some cases an exon capture specific for TK genes followed by massive parallel sequencing was applied [81-84]. Since these analyses can be performed on formalin-fixed paraffin-embedded samples, they can be applied to routine clinical practice. The possibility to test archived material will also allow for the characterization of the clonal evolution of TKF-bearing tumors, with consequent implementation of cancer tailored therapies.

Clinical diagnostic tests: tumor fingerprints

Clinically, TKF in human hematological cancers can be detected by an array of techniques. Genomic/conventional cytogenetics (CG) or FISH are commonly used for the detection of BCR-ALB translocations. Negative samples (<5% of cases) frequently carry variant translocations or cryptic rearrangements, which can occasionally be detected by FISH. Both approaches are also used to monitor drug responses and disease evolution. FISH and RT-PCR have shown higher sensitivity at diagnosis and better MRD detection than CG. In particular, qRT-PCR has become the gold standard for the diagnosis of MRD [49]. FuseFISH probe-based approaches for the detection of fusion transcripts have also recently emerged [85], but their usefulness in clinical practice is uncertain.

Resistance to TKi treatments is currently detected by RT-PCR coupled with DNA Sanger sequencing, but new methods have been developed. Massive parallel sequencing is rapidly replacing the above reported strategies, as it can also identify subclonal populations and characterize tumor clonal heterogeneity [86]. Analogous approaches have now been extended to the detection of actionable mutations, copy number variations, and gene rearrangement in clinical cancer specimens.

In solid tumors, TKFs are clinically identified by several techniques, some of which are performed as companion tests to select the best TKi agent. In 2013, the USA and China's FDA approved FISH- and IHC-based diagnostic tests to identify ALK fusions or ALK protein expression in NSCLC. Similar strategies were proposed to assess ROS1 and RET fusions [87]. Alternatively, a NGS-based detection kit for RET-rearrangements has been used to enroll patients in NSLCL clinical trials (Eisai sponsored trial; NCT01829217). It is conceivable that screening approaches to detect simultaneously multiple fusions will be implemented in the clinical arena [88].

Since different TKFs can be present in distinct cancer subsets, multi-targeted platforms have emerged as powerful tools to screen diagnostic samples. As multi-gene screening methods are entering into routine diagnostic procedures, new technical and therapeutic questions need to be addressed. Indeed, the greater efficacy of TKi compared to conventional treatments needs to be established for naïve patients carrying targetable fusions. Similarly, patients treated with multiple rounds of chemotherapeutic regimens need to be evaluated for the compliance to selective TKi, as limited data are available to justify their usage.

TUMOR-ASSOCIATED TK TRANSLOCATIONS: CURRENT THERAPEUTIC STRATEGIES

The discovery of recurrent tumor-associated TK translocations has brought new strategies for the treatment of hematologic and non-hematologic tumors. In the last decades, TKi have revolutionized the treatment of CML, Ph-positive B-ALL [89] and NSCLC carrying ALK or ROS-1 gene rearrangements [14, 90].

TKi-based therapies for the treatment of hematological tumors

BCR-ABL1 gene rearrangements are reported in all cases of CML and in about 30% of adult B-ALL. The fusion gene usually results from the t(9;22)(q34;q11.2) translocation, but 5-10% of cases bear variants and/or cryptic translocations [2]. BCR-ABL1 oncogenic properties are due to the constitutive activation of ABL1 TK [91, 92] and the loss of ABL-1 N-terminus regulatory domain.

Clinically, Imatinib has revolutionized the prognosis and outcome of CML, converting a highly lethal hematological neoplasm to a chronic and curable disease [49]. Complete cytogenetic responses (CCyR) are indeed reported in 87% of Imatinib-treated patients, with a progression-free survival (PFS) of 93% [93]. After 2 years of sustained CCyR, the life expectancy for CML patients is comparable to that of the general population [94]. Despite these results, several studies pinpointed a subset of cases with primary/secondary resistance to Imatinib, which can be overcome by new generation TKi (i.e. Nilotinib, Dasatinib and Bosutinib) with improved clinical efficacy [95]. These lines of evidence led the European Leukemia Net to recommend Imatinib, Nilotinib and/or Dasatinib as first line drugs for the treatment of CML. More aggressive therapeutic approaches (e.g. hematopoietic stem cell transplant) are considered only in very selected cases (e.g. refractory/relapsing cases unresponsive to TKi or accelerated/blast phase CML) [49].

In CML patients, resistance to TKi can result from two major mechanisms: (i) occurrence of TKi-resistant BCR-ABL1 mutations; (ii) enhancement of BCR-ABL1-independent pro-survival pathways. BCR-ABL1-dependent resistance has been overcome by third generation TKi that also target the most common ABL1 mutations (e.g. BCR-ABL1T315I). The most promising therapeutic agents are the conventional TKi, Ponatinib, and the BCR-ABL1 switch control inhibitor, DCC-2036 [96]. Notably, good results with Ponatinib have been obtained even in accelerated and blast phase CML [97], while the distinct action of DCC-2036 may overcome pre-existing mutations, including T315I [96]. Pemovska et al. have recently re-proposed the utility of Axitinib, an anti-VEGFR molecule approved for renal cell carcinoma, to counteract the resistance phenotype of T315I gatekeeper CML. Indeed, Axitinib tightly binds the T315I mutated kinase and inhibits leukemic cell growth, showing a better toxicity profile than other new-generation drugs, such as Ponatinib [98].

BCR-ABL1-independent resistance is mainly due to the activation of alternative signaling pathways promoting neoplastic cell survival [99]. Such pathways can be either extrinsic (i.e. activated by the marrow microenvironment) or intrinsic (i.e. due to microenvironment-independent mechanisms). Both intrinsic and extrinsic pathways lead to the activation of STAT3, through its phosphorylation at Tyr-705 [100]. Combined inhibition of STAT via BP-5-087, a potent and selective STAT3 SH2 domain inhibitor and BCR-ABL1 leads to synthetic lethality in resistant CML and constitutes a very promising therapeutic approach for refractory/relapsing CML [100]. Another therapeutic opportunity is represented by glitazones, anti-diabetic drugs that down-regulate STAT5 expression by inhibiting the peroxisome proliferator-activated receptor-γ (PPARγ). Recent data show that their use may lead to the complete eradication of leukemic stem pools, thus greatly improving the survival of CML patients in sustained CMR [101].

Unlike what has been observed in CML, the use of single-agent TKi in Ph-positive ALL has not produced sustained clinical responses. This is probably due to the rapid development of TK mutations, which are reported in >80% of adult Ph-positive ALLs at relapse [102]. Ph-positive ALL and CML also greatly differ in terms of genomic backgrounds, which may influence the response to TKi. For instance, INK4-ARF deletions, which often occur in acute Ph-positive ALL with aggressive clinical course, facilitate the emergence of BCR-ABL drug-resistant clones. This molecular event is not observed in indolent CML or in CML in blast crisis [103].

These data provide the rationale for the use of multi-modal regimens for the treatment of Ph-positive ALL, based on a combination of intensive chemotherapy and TKi. Very encouraging results have been first obtained in pediatric trials (COG, EsPhALL and SHOP-2005 trials) [104, 105] and, more recently, in adult patients [106]. In particular, a recent Phase II trial on Imatinib mesylate plus HyperCVAD (Cyclophosphamide, Vincristine, Doxorubicine, Dexamethasone) has reported complete clinical remission in 93% of adults with active Ph-positive ALL, with a 5-year overall survival (OS) of 43%. Notably, allogeneic stem cell transplant (ASCT) after Imatinib and HyperCVAD improved median OS only in patients with residual molecular disease. The role of Dasatinib and Nilotinib for the treatment of pediatric and adult Ph-positive ALL is still under evaluation, but preliminary results indicate their safety and efficacy when used with high dose chemotherapy and/or ASCT [107]. Encouraging data have also been obtained with Dasatinib and corticosteroid, which proves the potential efficacy of chemo-free protocols [108]. Finally, the association of TKi and bispecific T-cell engaging antibodies (BiTE) or chimeric antigen receptors T-cells (CARTs) represents a promising strategy for the treatment of Ph-positive ALL [109].

TKI-based therapies for the treatment of solid tumors

The introduction of TKi for the treatment of ALK-, ROS1- and RET-translocated NSCLC [14, 110, 111] has led to an astonishing improvement of patients' outcome and survival. More recently, fusions involving NTRK1 and FGFR1/2/3 have been reported (Table 1), thus enlarging the spectrum of promising therapies for NSCLC (i.e. Crizotinib, Ceritinib, Alecitinib for ALK-positive NSCLC; Crizotinib for ROS1-positive NSCLC; Cabozantinib and Vandetanib for RET-positive NSCLC).

Crizotinib represents the paradigm of efficient translation of molecular knowledge to the bedside. Initially developed as a c-Met inhibitor, Crizotinib showed clinical activity on advanced ALK-positive NSCLC in both the PROFILE 1001 (phase I) and PROFILE 1014 (phase III) clinical trials. Of note, in such tumors, Crizotinib displayed even greater efficacy than standard chemotherapy [112]. Similar results have been observed in ROS1- or RET-positive NSCLCs, with good responses to Crizotinib in ROS1-positive cases [14] and to Cabozantinib in Crizotinib-refractory cells [113]. The application of TKi has recently expanded to RET-mutated medullary thyroid carcinoma, RET-translocated NSCLC and other solid tumors (NCT02183870, NCT0194502, NCT01823068, NCT01639508). More recently, inhibitors with broad spectrum of activities were tested (RXDX-101), providing novel strategies to target different molecular defects (NTRK1/2/3, ROS1 and ALK translocation) shared by wide range of solid tumors (NCT02568267, NCT02097810).

Despite these outstanding results, TKi often lose their efficacy due to the development of treatment-related resistance [114]. Much like hematologic neoplasms, primary and secondary mechanisms determine resistance to TKi in solid tumors. Primary resistance is linked to the tridimensional structures of the protein that confer TKi-resistance [115], whereas secondary resistance mainly consists of gene mutations, bypass mechanisms, and/or gene amplifications overcoming the effect of TKi at therapeutic levels.

As for target mutations, different genetic alterations associated with resistance have been described for ALK-positive NSCLC: i) L1169M within the “gatekeeper” domain (the most frequent); ii) L1152R, associated with EGFR signaling activation; iii) G1202R and S1206Y within the ATP-binding pocket, leading to a decreased Crizotinib affinity [116]. The most frequently reported bypass mechanisms associated with secondary resistance to ALK inhibition are: i) EGFR/KRAS mutations [117] and Tyrosine Kinase Receptor (ERBB, c-MET, etc.) activation triggered by paracrine stimuli [118, 119]; ii) c-KIT amplification, requiring SCF [120], IGF-1R up-regulation [51] and recruitment of P2Y receptors [121]. Other resistance mechanisms include impaired drug influx (OCT-1), increased drug efflux (MDR1), or microenvironment-mediated resistance [122].

Different strategies have been proposed to overcome TKi acquired resistance, including the development of next-generation drugs, such as the ALK-inhibitors LDK378-Ceritinib and AP26113-Brigatinib (for ALK-positive NSCLC resistant to Crizotinib), Lecitinib (for ALK-positive NSCLC carrying the L1169M mutation) and PF06463922 (a dual ALK/ROS1 inhibitor with a good CNS penetration profile) [123]. Additionally, Hsp-90 inhibitors (AUY922-Ganetespib) are currently under evaluation in phase I/II clinical trials. Other emerging strategies include the administration of combined therapies to by-pass the “addiction” to single TKFs (e.g., combined inhibition of HER family [Dacomitinib], c-KIT [Dasatinib], src [Saracatinib], MEK [AZD6244], and IGF-1R [Linsitinib]). The re-timing of drug administration should also be considered, as recently demonstrated by the case of a Crizotinib-resistant ALK-positive NSCLC, displaying good response to this TKi after a withdrawal period [124]. Similar results have been obtained in an EGFR-mutated NSCLC, which regained sensitivity to EGFR-inhibitors upon therapy discontinuation [125]. Lastly, immunotherapeutic strategies (i.e. vaccination against over-expressed oncogenic antigens) may represent an effective approach for solid tumors, as recently reported in ALK-positive disorders [59, 126].

A final major roadblock concerning TKi is the poor response of central nervous system (CNS) metastases to therapy. This drawback may be overcome by different strategies: (i) combined therapies with drugs favoring TKi CNS accumulation, as recently attempted through the co-administration of Vandetanib and Everolimus (an immunosuppressive drug also modulating the P-glycoprotein efflux activity) [127]; and (ii) usage of single compound with higher brain penetrance, such as Ceritinib compared to Crizotinib.

FUTURE PERSPECTIVES

Despite TKi have dramatically changed the clinical management of numerous cancers, many questions have still to be answered. One of the major problems concerning TKi is the high frequency of primary refractoriness/tumor relapses [128] (Figure 2). Possible solutions for this limitation are: (i) the development of more potent drugs with limited side effects; (ii) the co-administration of TKi and chemotherapeutic agents [129], immune-based protocols [130] and/or bone marrow transplant [131]; and (iii) further research on treatment timing and dosage.

Figure 2

Mechanisms of Intrinsic and/or Acquired Resistances to Tyrosine Kinase Exposure

Multiple mechanisms are associated with the emergence of drug resistance. These include: development of secondary mutations in the Kinase Domain (KD) at gatekeeper sites; copy number gains; activation of alternative oncogenic pathways via somatic mutations (i.e. RAS), and compensatory “by-pass” routes by Receptor Tyrosine Kinases signaling (i.e. EGFR, HER-2, HER-3, c-MET). Alternatively, resistance may also be due to either paracrine external signals through the tumor microenvironment, or via autocrine loops. The major drug categories and their therapeutic modalities are indicated.

Of note, a growing number of protocols combining chemotherapy and TKi are currently being tested in both TKF-induced hematological and solid tumors [107, 132] [133], and several trials are evaluating the efficacy of new TKi either alone or in combination with more conventional therapies (NCT02269085; NCT02427620; NCT02315768; NCT01606878; NCT02134912; NCT02511184). Preclinical studies have also demonstrated a substantial advantage of combining classical chemotherapy with TKi: Das and colleagues showed a synergistic effect of Crizotinib and Temozolomide in FIG-ROS1-positive glioblastoma patient-derived cells, paving a route to improve clinical outcome [134]. Similarly, Appelmann and colleagues evaluated the possibility of combining Dasatinib, Ruxolitinib and Dexamethasone to prolong remission and avoid major toxicities in a mouse model of Ph-positive ALL [135].

One of the most challenging issues on the combined use of TKi and chemotherapeutics concerns the optimal doses and schedules to use. Toward this end, mathematical models might improve the clinical success, limiting unacceptable side effects and TKi/conventional therapy-resistance [136]. This approach also needs dedicated trials, the cooperative agreement of pharmaceutical companies and new regulations on intellectual and commercial issues. Moreover, we need to develop biological and mathematical models to predict cancer evolution, capable of predicting cancer evolutionary trajectories based on pre-treatment diversity. Lastly, we foresee that these predictions need to be adjusted in each individual patient by integrating longitudinally acquired molecular readouts representative of the global cancer landscape (i.e. comprehensive genomic data [WES] of multiple sites, liquid biopsies, etc.).

Other questions regarding TKi concern the risks and benefits of second/third generation drugs. Thus, the use of first generation compounds remain a valuable option in many settings, including highly targeted diseases and the management of third world populations. The increasing health costs, the uncontrolled fee schedules in many industrialized countries, the increasing co-payer costs and the accessibility of generic TKi represent very tangible issues which may be solved only via international based supports and regulations. Finally, the need of cutting health system costs poses unprecedented challenges to molecular medicine programs. In fact, the success of such therapies depends on the characterization of individual cancers (both primary and recurrent/metastatic lesions). These analyses have to be performed by a pool of medical centers following certified programs, with consequent huge health costs.

These practical problems are paralleled by a series of scientific and methodological challenges. Although the genomic characterization of each cancer has originally prompted the use of selective compounds, it is now clear that this approach may not be sufficient. Indeed, favorable responses to selective drugs are not strictly linked to defined genetic defects. Thus, functional tests (e.g. comprehensive phosphorylation maps) should be performed together with genomic analyses. As an alternative, in vitro preclinical tests on patient-derived cell lines or PDTX may be introduced (Figure 3).

Figure 3

Comprehensive Management of Tyrosine Kinase-driven Cancers

Molecular and functional characterization of human cancers in combination with drug screening tests on primary patient tumor samples is expected to drive precise target therapies. It is anticipated that tumor samples will be used to generated 2D/3D and well as PDTX models. In vitro models from primary samples or from PDTX will serve to screen large libraries of compounds, whose efficacy will be established using a battery of biological and molecular readouts. The data emerging from these high throughput screenings will be pivotal to test selected molecules in preclinical PDTX-based trials. Extensive molecular and functional readouts will then be obtained from both tumors and host compartments (i.e. plasma/liquid biopsy) to create a detailed profile of in vivo classifiers and biomarkers. These will ultimately serve as surrogates to predict and define response to specific therapies in cancer patients. The development of tumor biorepositories of primary and metastatic lesions, cell lines and PDTX from primary human cancers is predicted to lead to the recognition and understanding of new oncogenic events. It is anticipated that effective targeted therapies will improve clinical responses and ultimately will lead to lower health care costs.

Of note, in vitro tests with patient-derived tissue samples may allow personalized pharmacologic interventions without the limits related to the use of immortalized cell lines or other tumor models. Patient-derived cancer models are indeed pivotal for drug discovery, for the characterization of cancer biology and for the assessment of personalized therapies [137-139]. Several cancer PDTX models have been developed, but only handful of them carries TKFs [15, 140-142]; thus, a more systematic effort in generating “ad hoc” models is highly recommended.

We anticipate that in vitro primary cell lines will allow for the rapid testing of drug libraries, with subsequent validation in more complex models (e.g. PDTX). It is also desirable that new clinical trials will include parallel co-clinical mouse PDTX-trials, which will establish the efficacy of such models. The molecular stratification of these models will finally provide new cancer classifiers with possible diagnostic/prognostic implications (Figure 3).

The considerable therapeutic progress against TKF-driven tumors may also prove beneficial for the treatment of those translocation-negative neoplasms, which nonetheless carry defects of TKF-related signaling pathways. As previously highlighted, TKFs indeed represent only one of the several mechanisms leading to TK dysfunction, and other TK impairments (i.e. gene mutations/amplifications or non-fusion rearrangements) lead to phenotypes closely resembling those of TKF-driven neoplasms. This is, for instance, the case of subsets of ALK-negative ALCL, carrying JAK1 and STAT3 activating mutations, which induce transcriptional signatures akin to those of ALK-positive ALCL [16]. Moreover, it is now recognized that activating mutations can synergize with TFKs and lead to full neoplastic transformation [8]. This molecular evidence constitutes the rationale for the administration of TKi compounds in specific subsets of both TKF-negative and TFK-positive tumors. Further studies and clinical trials will help clarify this captivating opportunity.

In conclusion, it is desirable that the academic community promotes dedicated programs on TKi and TKF-bearing cancers, also encouraging the support of pharmaceutical companies and the distribution of compounds under compassionate use. This approach, together with the creation of open-access clinical databases, will greatly contribute to the understanding of cancer-related TKFs and to a better management of (TKF-positive and TKF-negative) cancer patients.