RTKs and criteria for rational therapeutic targets
We will now consider alterations specific to RTKs. The most valuable therapeutic targets will be those providing critical transforming signals that drive tumour growth and survival, where inactivation results in selective apoptosis and tumour regression. This dependence has been termed ‘oncogene addiction’ (Weinstein and Joe, 2008). Oncogene addiction is seen for several tyrosine kinases, such as BCR-ABL in chronic myeloid leukaemia, targeted by imatinib (Druker, 2004; George et al., 2004). Targeted therapy has been most effective when directed against tumours expressing highly activated kinases, such as BCR-ABL in acute myeloid leukaemia and EGFR in NSCLC. Pointers indicating a good RTK therapeutic target in a cancer type include reports of the following: (i) high expression, (ii) activating mutations, (iii) tyrosine phosphorylation (indicating RTK activation) and (iv) inhibition of RTK activity in functional studies, with apoptosis and growth arrest. In addition, the target should lack a vital role in normal adult physiology (Bauman et al., 2007; George, 2002). Here, we review findings concerning RTK sequence, expression, activation and function in melanoma. Five RTKs that are, or may become therapeutic targets, are then considered in further detail.
Sequence variants of RTKs in melanoma
Until around 2007, there was surprisingly little evidence for driver mutations within RTKs in melanoma, although these were long expected, because RTK activation would upregulate all the pathways in Figure 1. Downregulation or loss of the vital melanocytic RTK, KIT, had been reported in most cutaneous melanomas, as reviewed previously (Easty and Bennett, 2000). However, activating KIT mutations have now been identified in some melanoma subtypes (mucosal, acral and non-sun-exposed) in which BRAF and NRAS mutations are rare; here, KIT is still expressed (Ashida et al., 2009; Curtin et al., 2006). FGFR1-activating mutations were likewise reported in two melanomas lacking activated RAF or RAS (Thomas et al., 2007).
Technological advances have now allowed a comprehensive cataloguing of somatic mutations within the large PTK gene family (kinome), in cancers including human melanomas and cell lines derived from them. (Note: ‘mutation’ in these studies designates any non-conservative sequence alteration not seen in the germline, and may have included some changes not affecting function.) All 86 PTK genes were sequenced in established tumour cell lines including 53 melanoma lines and made available as a database of tyrosine kinome variants (Ruhe et al., 2007). In general, there was a low incidence of these somatic mutations; still, mutations were identified within 19 RTKs including MET, TYRO3, EPHA2, EPHB2, EPHB6 and NTRK1-3 (summarized in Table 1). Mutated cytoplasmic PTKs included FAK and PTK2B, both in the FAK subfamily (Ruhe et al., 2007). In a second large-scale study, the 86 PTK genes were sequenced either partially or fully from 79 melanoma cultures (Prickett et al., 2009). Again, they described somatic mutations in 19 PTK genes, but largely different from those seen by Ruhe et al. (2007). Mutations in receptors are listed in Table 1. They include notably frequent somatic mutations in ERBB4 (19%); also EPHB2, EPHB6 and VEGFR1 (9-10% each). Cytoplasmic PTKs with mutations included PTK2B and FAK again; also FER, PTK6 and PTK7.
Table 1. Receptor tyrosine kinases (RTKs) in melanoma: expression changes, mutation and activation in cultured cells
|TYRO3|| Re, Nf||(+)d||100|
|MER|| || ||+|
|DDR||DDR2|| Ng,f|| || |
|ERBB2|| || ||53|
|ERBB3|| M, Ii||(+)d|| 26|
|EPH||EPHA1|| || ||0|
|EPHA2|| Nk, Wg,l, Ig,m, Rl||(+)d||0|
|EPHA4|| Wn, M||(+)d||+|
|EPHA7|| || ||0|
|EPHB2|| ||8.9%j; 5%d||93|
|EPHB3||Nf (94%)|| || |
|EPHB4|| || ||0|
|EPHB6|| Io, Ro, M||8.6%j, 5%d||66|
|FGFR||FGFR1|| R, Ip,q||(+)r 5%d||0|
|FGFR3|| || || 60|
|FGFR4|| Ng,f|| ||0|
|INSR||INSR|| || ||33|
|IGF1R|| It, M|| ||47|
|MET||MET (HGFR)|| I, Wu||(+)u, 5%d||0v|
|MST1R (RON, MSP-R)|| ||(+)d||33|
|MUSK||MUSK|| || ||0|
|PDGFR||PDGFRA|| Iw M||5%j||+|
|PDGFRB|| || ||0|
|KIT|| Wx,y, Nx,z, Iq,aa||29%bb, 18%aa||0|
|FLT3 (FLK2)|| ||(+)d||+|
|CSF1R (MCSFR, FMS)|| Icc|| ||0|
|PTK7||PTK7 (CCK4)|| Nn,f|| || |
|RET||RET|| || ||+|
|ROR2|| || ||0|
|RYK||RYK|| ||(+)d|| |
|TEK (TIE2)|| ||(+)d||+|
|TRK||NTRK1 (TRKA)|| ||2.5%j, 5%d||+|
|NTRK2 (TRKB)|| M||(+)d||0|
|NTRK3 (TRKC)|| M||(+)d||0|
|VEGFR||FLT1 VEGFR1)|| Mk||10%j (+)d||0|
|KDR (VEGFR2)|| Mk, Nf||5%d|| 66|
|FLT4 (VEGFR3)|| Mk|| ||+|
The reason for the largely different findings between these two studies is unclear. Both groups studied metastatic, not primary melanomas. However, Ruhe et al. analysed established cell lines, comparing them to unmatched non-tumourigenic cell lines and normal tissues, while Prickett et al. used fresh, low-passage cultures, compared to matched normal genomic DNA from patients’ peripheral blood. Ruhe et al. sequenced all exons in full, while Prickett et al. sequenced the kinase domain of all PTKs, and then all exons of the 19 PTKs that showed kinase-domain mutations. Differences may also reflect the overall diversity of cancers and a paucity of very common PTK alterations. Some of the mutations are likely to play a role in melanoma progression (drivers); others may represent passengers (random, neutral changes). Functional studies of ERBB4 are discussed later. Mutations of PTK2B were found in a substantial fraction (10%) of melanomas. Identified in both studies, PTK2B is a member of the FAK subfamily. It binds to a GTPase regulator associated with FAK, and the SH2 domain of GRB2 (Schaller, 2010). Separate studies of specific RTK genes in melanoma describe mutations in MET, KIT and FGFR2, the last being inactivating mutations reminiscent of the common losses of KIT already mentioned (summarized in Table 1). A linking theme may be that where there is downstream oncogenic activation, of proteins such as BRAF, RAS or MITF, then upstream receptors may become redundant, or even inhibitory by pathway overactivation. Otherwise, RTKs may themselves become oncogenically activated and/or amplified.
Altered RTK expression in melanoma
Significant increases in expression of a given RTK in melanoma compared to melanocytes suggest that the kinase may promote tumour progression. RTK expression has been extensively studied in melanomas and precursor lesions. Previous studies used northern and immunoblotting analysis and immunohistochemistry. These were reviewed previously, when 12 RTKs and 4 non-receptor PTKs to date had been reported to be overexpressed or ectopically expressed relative to normal melanocytes. Some of the most common were EPHB3 (94% of melanoma lines), EPHA2 (93%) and FGFR4 (64%) (Easty and Bennett, 2000). More recently, microarrays and RT-PCR analysis of candidate genes have been extensively applied and permit further meta-analysis of melanoma RTK expression (Table 1). Microarray data in the Oncomine database identify apparently increased expression in uncultured melanomas of mRNAs for ERBB3 (3- to 5-fold increase) and IGF1R (2-fold increase) compared to normal skin, and TRKC (18-fold increase) compared to benign naevi. Expression of several RTK mRNAs (MSPR, NTRK2/TRKB and EPHA4) was lower in melanoma compared to naevi (Oncomine, http://www.oncomine.org/resource/login.html). Note: ‘abnormal’ gene expression in melanoma would be expected to mean different from that in the normal cellular counterpart, melanocytes. As melanocytes comprise only a small fraction of cells in normal skin, some studies (as found in Oncomine) compare melanoma with whole skin, which may give misleading results such as suggesting that melanocyte-specific genes are ‘overexpressed’ in melanoma. Naevi provide a better control than skin, but still have various differences from melanocytes. Thus, these comparisons should be interpreted with caution unless confirmed by other methods. Table 1 summarizes reported changes in RTK levels.
RTK activity and abnormalities in melanoma
Studies of RTK mRNA or protein abundance may not always reflect activation. More recently, phospho-arrays have been used to analyse RTK activity in melanoma (Margaryan et al., 2009). These employ arrayed antibodies specific for the phosphorylated (usually the active) form of a protein kinase, permitting simultaneous semi-quantitative analysis of receptor activity in cell or tissue extracts (Margaryan et al., 2009). Differing signal intensities on a phospho-RTK array may arise from the following: (i) differences in abundance, via differing transcription, translation or degradation; (ii) activating or inactivating mutations in RTKs, (iii) the presence of cognate ligands (exogenous or endogenous) and (iv) activity of PTPs or other interacting components. We used Proteome ProfilerTM Arrays (R&D Systems) to assess phosphorylation of 42 RTKs in 17 melanocytic cell lines of various degrees of malignancy. (See Supplementary Information for Materials and Methods.) Non-small cell lung cancer (HCC827) cells containing amplified, kinase-activated EGFR (36 copies) were used as a positive control and yielded an intense signal for EGFR as reported previously (Engelman et al., 2007).
Each cell line was tested on arrays in duplicate assays, with consistently reproducible results. Representative arrays are included as supplementary data. Figure S1 indicates grid orientation on the Proteome ProfilerTM Array. We analysed cells from various stages of melanoma progression, including normal melanocytes (Figure S2), radial growth-phase melanoma (Figure S3), vertical growth-phase melanoma (with data from co-isogenic melanoma cell lines, Figure S4), and metastatic melanoma (Figure S5). Although the Proteome ProfilerTM Arrays (R&D Systems) have been widely used to assay kinase activity within other tumours (Ball et al., 2007; Eckstein et al., 2008; Engelman et al., 2007; Stommel et al., 2007), data presented here remain preliminary and will require further validation by specific RTK immunoprecipitation and immunoblotting with anti-phosphotyrosine antibodies (MS in preparation).
A complex pattern of 25/42 activated RTKs was detected in melanocytic cells (Figure 2; summary in Table 1). Phosphorylation signal intensity for melanocytic cells was low to moderate: typically 30-fold less than the amplified EGFR control. Normal melanocytes showed some variation in activated RTKs, but EGFR, TYRO3, KIT, TIE1 and EPHB2 were consistently active in both lines. The KIT ligand stem cell factor (SCF) is an additive to the culture medium, EGF and macrophage-stimulating protein (MSP, the RON ligand) are present in serum, and it seems that ligands for TYRO3 (bovine protein S, see below) and for TIE1 must also be present. (Note that all cells were grown with serum.) TIE1 and TIE2 have been identified as receptors for angiopoietins (Seegar et al., 2010), vasculogenic regulators expressed by various cells including some tumours, but it is unknown whether they are present in normal melanocytes. TYRO3 is discussed later. Phosphorylation of EPHB6 was seen in a proportion of lines. EPHB6 lacks kinase activity, so this may reflect heterodimerization with an active RTK, with cross-phosphorylation of EPHB6.
Figure 2. Summarized activity of 42 receptor tyrosine kinases (RTKs) in 17 melanocytic cell lines. These were normal melanocytes (Nohm-1, 830c), RGP melanoma (WM35, WM1650, SGM2), VGP melanoma (ME1402, WM9, WM1341D, WM98.1, WM793, WM793P1), metastatic melanoma (1205LU, WM852, WM239A, WM1158, DX3, A375P) and a control NSCLC cell line (HCC827). All cells were grown in the presence of serum, which may have been the source of ligands for some RTKs. Signals with intensity >2 × SD above the mean intensity of 10 negative controls were scored as positive. Red shading indicates high and green shading indicates low or undetectable kinase activity.
Download figure to PowerPoint
None of 15 melanomas showed KIT activity. This included SGM2, grown in the presence of SCF, so that differences between cell lines are unlikely to reflect solely differences in composition of culture medium. TIE1 activity was lost in a proportion of melanoma lines, though without an obvious relation to progression. EGFR and TYRO3 activity were maintained in all melanomas and EPHB2 in all but one, suggesting a positive role. A number of RTKs were activated in at least 2 of the 15 melanoma lines but not in melanocytes, namely ERBB3 (four lines), FGFR2 (6), FGFR3 (9), INSR (6), PDGFRA (2), FLT3 (2), VEGFR2 (10), VEGFR3 (2) and EPHA4 (2). Three isogenic lines showed similar patterns: WM793 (poorly tumourigenic parental cells), WM793P1 (more tumourigenic derivative) and 1205Lu (metastatic derivative), with a few differences that may have been related to increasing malignancy but further study would be needed. Our initial expectation was that RTK arrays might identify kinases over- or underactive in advanced melanoma cell lines. However, no clear pattern of changing RTK activation emerged in early versus advanced melanoma in this study. Nor was any single RTK up- or downregulated in all melanomas. However, at the receptor family level, upregulated activity was seen in 12/15 melanomas for an FGF receptor and 11/15 for a VEGF receptor. The latter may be because of increased receptor expression, from mRNA studies (Table 1).