RET is a receptor tyrosine kinase (TK) expressed primarily on derived neural crest and urogenital cells during embryonic development. It is required for maturation of several cell lineages of the peripheral nervous system, kidney morphogenesis and spermatogenesis ² . The glial derived neurotrophic factor (GDNF) family of ligands bind RET in association with one of four glycosyl phosphatidylinositol (GPI) anchored GDNF family α-receptors (GFRα), triggering RET dimerization, followed by auto-phosphorylation of specific tyrosine residues within the C-terminal chain and trans-phosphorylation of intracellular signalling cascades. These downstream signalling networks play a key role in regulating cell survival, differentiation, proliferation, migration and chemotaxis ³ .

Activating mutations in RET ( e.g. C634W and M918T) have been identified in familial and sporadic forms of medullary thyroid carcinomas (MTC ^{4, 5} ) and are associated with aggressive disease progression ⁶ . More recently, several groups independently identified RET rearrangements in 1–2% of lung adenocarcinoma (LAD) cases ^{7– 10} . The RET fusion genes discovered in these studies include CCDC6-RET (already known as RET/PTC1 in papillary thyroid carcinoma) as well as a novel fusion with KIF5B (kinesin family member 5B), encoding a coiled coil domain, generated by pericentric inversion in chromosome 10. The coiled-coil domains present in the fusion partner promote overexpression and ligand-independent dimerization leading to constitutive activation of RET. These studies also demonstrated that the resulting fusion proteins are oncogenic, and that their inhibition has therapeutic potential. Importantly, the RET fusions are mutually exclusive with other known drivers in LAD (e.g. KRAS, epidermal growth factor receptor (EGFR), EML4-anaplastic lymphoma kinase (ALK)), further supporting a role for RET as a unique driver of malignancy in these tumors. RET-positive patients represent a well-defined population with specific features: all are adenocarcinomas, and patients tend to be non-smokers and to be younger than the median age for lung cancer patients ¹¹ .

At present, there are no known specific RET inhibitors in clinical development, although many potent inhibitors of RET have been opportunistically identified through selectivity profiling of compounds initially designed to target other TKs. The small molecule inhibitors vandetanib and cabozantinib are perhaps the best examples of such compounds. Although both have been approved for the treatment of advanced metastatic MTC ^{12, 13} , RET inhibition is a secondary pharmacology of these drugs, which were initially developed as inhibitors of other TKs. Both agents target KDR, whilst vandetanib has additional activity versus EGFR and cabozantinib versus MET. These compounds are now under investigation for the treatment of RET fusion positive LAD. A preliminary report of a phase II trial ¹⁴ of cabozantinib confirmed partial responses in two of three RET-positive patients ¹¹ ; the third patient presented with prolonged stable disease. The activity of vandetanib in RET fusion positive patients has been demonstrated in two case reports ¹⁵ . However significant toxicity ( e.g. rash, diarrhoea, hypertension) resulting from inhibition of other kinases, particularly KDR, has led to dose reductions in clinical trials (11–13) and is likely to compromise the use of both these agents in clinical settings ¹⁶ . Thus, there is a clear need for selective RET inhibitors which do not display the toxicities associated with the current treatments and enable more potent and sustained inhibition of RET signalling. These agents may offer greater clinical benefit for patients with RET mutant cancers and widen the scope for the clinical use of RET inhibitors ¹⁷ .

The role of RET in this subset of LAD has heightened interest in re-purposing a number of other clinically approved inhibitors, shown to have RET activity in pre-clinical studies. Sunitinib, sorafenib, ponatinib and lenvatinib, all multi-kinase TK inhibitors with some RET activity, are currently under investigation in numerous phase II clinical trials ¹⁴ for treatment of RET fusion positive LAD ¹⁸ . Sunitinib, already approved for the treatment of imatinib-resistant gastrointestinal stromal tumors (GIST), advanced renal carcinoma and advanced pancreatic neuroendocrine tumors, is the subject of a phase II study in certain types of LAD tumors, including those harbouring a RET fusion. Sorafenib, also approved for several indications including kidney and liver cancer, has demonstrated preclinical activity in RET models but has yet to be tested in patients selected based on RET fusion status. Some efficacy in advanced MTC has been reported for lenvatinib ¹⁹ , however tumor response did not correlate with RET mutation status and the observed toxicity profile was consistent with KDR inhibition. A phase II study of lenvatinib in RET fusion positive LAD is ongoing ¹⁴ . Ponatinib is also a multi-targeted, broad-spectrum tyrosine kinase inhibitor ²⁰ , approved in late 2012 for patients with resistant or intolerant chronic myeloid leukemia and Philadelphia chromosome-positive acute lymphoblastic leukemia. Ponatinib was withdrawn shortly afterwards due to serious safety concerns but was later returned to the market with additional warnings in the product information. An investigational phase II clinical trial of ponatinib in LAD patients selected based on RET mutation status is currently ongoing ¹⁴ . Alectinib is a highly selective ALK inhibitor (median inhibitory concentration of 1.9 nM for ALK activity), recently approved by the FDA for the treatment of patients with ALK positive LAD who progressed on crizotinib. Preclinical data demonstrating activity of alectinib in RET mouse models ²¹ has led to its investigation as a treatment for RET fusion positive LAD as part of the DARWIN II trial ¹⁴ . The compounds described above are all classic TK inhibitors but in 2010, Alfano et al. ²² proposed an alternative approach for targeting RET in LAD involving inhibition of HSP90 (heat shock protein 90kDa). HSP90 is a molecular chaperone that plays a central role in regulating the correct folding, stability and function of numerous proteins ²³ . Inhibition of HSP90 activity results in aggregation or proteasomal degradation of these proteins. RET, along with other driver kinases such as EGFR and ALK is a HSP90 client protein and as such requires HSP90 for protein stability and function. Thus, targeting the chaperone function of HSP90 offers an alternative to direct kinase inhibition for therapeutic intervention in RET driven cancer. Clinical evaluation of this approach is currently being assessed as part of a larger study in stage IV LAD patients with driver molecular alterations other than EGFR mutations ¹⁴ .

In an earlier publication ¹ we reported the discovery of 2-substituted phenol quinazolines as potent RET kinase inhibitors with improved KDR selectivity; here we describe development of the robust screening cascade which allowed us to achieve that goal and additionally profile existing clinical compounds with RET activity in order to assess them against our compound target profile.

Our aim was to develop and validate a robust screening cascade to support the identification and development of potent and selective RET inhibitors using vandetanib as the starting point ¹ and to assess these attributes in clinical compounds reported to have RET activity. All compounds were initially assayed biochemically for activity versus RET and KDR enzyme. Once potency and selectivity had been confirmed, and structure activity relationships (SAR) demonstrated, for a number of compounds within the anilinoquinazoline series ¹ , we performed further biochemical studies to investigate the mechanism of RET inhibition using selected compounds including our starting point, vandetanib ( Figure 1). Some inhibitors bind to, or dissociate from the target enzyme slowly, leading to time dependent inhibition ²⁵ . Failure to identify this can lead to an underestimation of biochemical potency and misleading SAR. To investigate this, compounds were assayed under standard conditions but pre-incubated with RET for between 0 and 60 minutes prior to addition of ATP. Pre-incubation did not significantly affect IC ₅₀ values indicating that the anilinoquinazolines are not slow binders and that a 15 minutes pre-incubation, as used in our standard assay, is sufficient to allow the reaction to reach equilibrium ( Figure 1A). Although diverse in primary amino acid sequence, the human kinases share a great degree of similarity in their 3D structures, especially in their catalytically active kinase domain where the ATP-binding pocket is located. Kinase inhibitors can be grouped into two classes, based on binding mode: irreversible and reversible. The former tend to bind covalently with a reactive nucleophilic cysteine residue proximal to the ATP-binding site, resulting in the blockage of the ATP site and irreversible inhibition. Reversible inhibitors can be further classified into four main types, competitive, non-competitive, uncompetitive and mixed inhibition ²⁶ . Our data ( Figure 1B) indicates that the anilinoquinazolines under investigation here fall into the reversible, ATP competitive category which is not surprising since it has been demonstrated previously that vandetanib, a related compound and our starting point, exhibits this mode of binding ²⁷ .

Compounds meeting defined criteria for potency and selectivity in the biochemical assay were then assessed in a cellular POM assay. Initially we developed an ELISA-based POM assay, measuring changes in levels of (active) phosphorylated RET (M918T) and KDR, within the same MTC cell line (MZ-CRC-1).

However, it soon became apparent that compared to KDR there was a poor correlation between RET biochemical and cell-based IC ₅₀ measurements ( Figures 2A & B). In addition, the drop off in potency from the biochemical to cellular assay was generally much greater for RET than KDR, in effect compressing the selectivity margins, often from >100-fold in the biochemical assay to parity (or worse) in the cellular assay. Compound permeability was not believed to be the cause of the disparity as the magnitude of the drop off was not the same for both RET and KDR, despite measurement of both proteins in the same cell lysate. A previous report ²⁸ had indicated that a single mutation within the kinase domain of RET could alter compound inhibitory activity. In order to address the possibility that some of the discrepancy could be due to our measurement of WT protein activity biochemically but mutant RET (M918T) protein in the MZ-CRC-1 cells, we compared the biochemical activity of selected compounds versus both RET (WT) and RET (M918T) protein. The compound IC ₅₀ data for the two proteins correlate well ( Figure 2C) and therefore indicate that the mutation does not significantly affect biochemical activity for this class of compound. In order to follow this up in the cellular context we extended our repertoire of assays to enable robust measurement of RET activity in TT ²⁹ , LC-2/ad ^{30, 31} and mouse BaF3 ¹⁰ cells harbouring a C634W mutation, a CCDC6-RET fusion and a KIF5B-RET fusion respectively. We assessed relative RET potencies for a number of tool compounds (including vandetanib and cabozantinib) across these cell lines and found that activities correlated well ( Supplementary Figure 1), further supporting the notion that this class of compounds are equally active versus mutant and WT forms of the protein.

In order to demonstrate POP in a disease relevant cell line, we developed and validated a proliferation endpoint assay in the MTC line MZ-CRC-1; non-specific toxicity was evaluated by measuring the same endpoint in HEK293, a human embryonic kidney line which does not express RET. Compounds exhibiting RET potency (<500nM) and selectivity (>10x) in the POM assay were selected for POP assay screening. Our data show that there is a good correlation between the POM and POP IC ₅₀ values measured in the MZ-CRC-1 ( Figure 3A) cells and other RET models (e.g. LC-2/ad, Supplementary Figure 2) indicating that the RET inhibitory activity that we observe in our POM assay translates into anti-proliferative effects in our MTC and LAD disease models. In addition, we have assayed a selection of representative compounds for anti-proliferative effects in LC-2/ad and as for the POM assay, data correlate well with that obtained in MZ-CRC-1 ( Figure 3B). Therefore, we are confident that activities observed in our routine MZ-CRC-1 POM and POP screens are predictive for the other cell line models.

Although at this stage we were confident that we could drive our internal chemistry effort based on the cellular data, it was clear that performance in the biochemical assay was not always a good predictor of RET cellular activity and therefore should not be used as a pre-screen for this chemical series. To address this issue this we developed and validated a higher throughput POM assay allowing parallel assessment of all compounds in both cellular and biochemical assays. This BaF3 proliferation assay platform, employing cell lines dependent upon RET or KDR for survival alongside WT IL-3 dependent control cells, is robust with 4x higher throughput than the original MZ-CRC-1 POM ELISA. More importantly the proliferation IC ₅₀ for both RET ( Figure 3C) and KDR in the BaF3 models correlate well with those generated using the respective phosphorylation endpoint assays thus maintaining the selectivity values previously observed ( Figure 3D). Accordingly, we introduced this as the routine POM screening assay with the option to further evaluate compound effects on phosphoprotein levels using the ELISA POM if necessary.

Chemical optimisation of the series ultimately led to the identification of a number of potent and selective RET inhibitors ( e.g. PDD16860 and PDD16964; Table 1). Clinical RET compounds currently under investigation in LAD were also profiled using the screening cascade ( Table 1). With the possible exception of alectinib, which it was not possible to dose at the highest concentrations due to solubility issues, none of the kinase inhibitors tested met our criteria for selectivity versus KDR ( i.e. >10x). Several reports in the literature ^{21, 32} , which we have subsequently confirmed in-house (data not shown), demonstrate that alectinib is a much better cellular inhibitor of ALK (IC ₅₀ in H2228 cells = 10nM) than RET (IC ₅₀ in MZ-CRC-1 cells = 80nM). Given this, along with its limited solubility and required dosing of 300–600mg bid for clinical efficacy in ALK driven LAD, therapeutic inhibition of RET is unlikely at clinically relevant doses. Although not a direct inhibitor of RET activity, we have also profiled ganetespib, a HSP90 inhibitor. Ganetespib did reduce levels of active RET, through protein degradation ( Supplementary Figure 3) but was not RET-selective and exhibited more toxicity (toxicity margin = 0.7x) in the control non-RET driven HEK293 cells compared to the MTC RET-driven model, MZ-CRC-1. This indicates that the pleiotropic effects of HSP90 inhibition on numerous client proteins (including KDR) is likely to elicit non-specific toxicity at therapeutically active doses. Other aspects of the screening cascade ( Figure 4) are already in place and as discussed in an earlier publication ¹ , selected compounds have been assessed in in vitro DMPK assays as part of the chemical optimisation. Furthermore, in vivo models and capabilities ( Figure 4) have been established and validated (data not shown) and several tool compounds are currently undergoing assessment in PK/PD and efficacy studies. To date, our robust screening workflows have proved very successful in identifying potent, selective tool compounds and will continue to support our active pursuit of clinical candidates suitable for evaluation in RET fusion positive LAD patients.

We have established a robust screening cascade to enable the identification and development of a clinical candidate compound demonstrating potent and selective RET activity. Furthermore we have profiled clinical compounds currently under investigation for treatment of LAD confirming that they do not meet our target profile. We are continuing our efforts to develop a clinical candidate and will report on further progress in a future publication.

The data referenced by this article are under copyright with the following copyright statement: Copyright: � 2016 Watson AJ et al.

Data associated with the article are available under the terms of the Creative Commons Zero "No rights reserved" data waiver (CC0 1.0 Public domain dedication).

F1000Research: Dataset 1. Datasets: selective inhibitors of RET and comparison with current clinical candidates through development and validation of a robust screening cascade, 10.5256/f1000research.8724.d122280 ³³