Insulin-like growth factor (IGF) promotes growth and mediates metabolic signals (1). There is evidence that links IGF to cancer or tumorigenesis, and examination of the literature (2–5) from the last decade suggests that agents capable of inhibiting IGF receptor would have potential in the treatment of such malignancies (6). To date, the first results from clinical trials on antibodies or small molecules targeting IGF-1 receptor are starting to be reported and confirm these hypotheses (7). We now wish to report our own findings in this area.
We initiated a high throughput screen (HTS) on IGF1-R using an Homogeneous Time-resolved Fluorescence (HTRF) assay relying on the receptor autophosphorylation. This led to the identification of a very potent hit 1 belonging to an azaindole chemical series from a Syk kinase inhibitor program (8).
As 1 was already quite potent on the target, we decided to evaluate it in an extensive panel of biological, Absorption, Distribution, Metabolism, Elimination (ADME), drug metabolism and Pharmacokinetics assays to anticipate potential liabilities and set the objectives of our optimization program.
We first evaluated 1 in a series of secondary assays related to IGF function. The potency in an ELISA IGF1-R autophosphorylation assay was comparable to the one obtained in the HTRF assay. Compound 1 was also able to inhibit the IGF1-induced proliferation in two different cell lines, a Mouse Embryonic Fibroblasts (MEF) cell line engineered to over express IGF1-R and an MCF7 breast tumor cell line, with IC50‘s of 1.6 and 4.3 μm, respectively (Table 1).
|IGF1-R (HTRF) (μm)||IGF1-R (ELISA) (μm)||IGF1-induced proliferation (MEF) (μm)||IGF1-induced proliferation (MCF7) (μm)|
Selectivity against a panel of kinases was also evaluated. As expected because of its origin, 1 was a potent Syk inhibitor (Table 2). As expected also because of its high homology, there was no selectivity versus insulin receptor kinase (IRK). Inhibitors of IRK would be expected to induce insulin resistance (hyperglycemia). It was hypothesized that for a short treatment period (about 4 months), there should not be any major metabolic disturbances (compared with the physiological insulin resistance during late state of pregnancy). In addition, some dual inhibitors of IGF1-R and IR-A such as 1H-Benzoimidazol-2-yl)-1H-pyridin-2-one BMS-536924 (9) are undergoing preclinical studies and could be advantageous for the treatment of tumours expressing both IGF1-R and IR-A (10). Finally, some inhibitors in development, such as the pyrrolo[2,3-d]pyrimidine NVP-AEW541 (11) displayed good selectivity for IGF1-R versus IR in intact cells in spite of no enzymatic selectivity for the kinases. Owing to the little experience at that time with non-selective IGF-R inhibitors, it was decided to assess the possible drawbacks of an unselective inhibitor later in the development. From this initial kinase panel, only KDR and GSK3β were inhibited albeit at higher concentrations (3.6 and 5.5 μm, respectively).
|Kinase||IC50 (μm)||ATP (μm)a||Kinase||IC50 (μm)||ATP (μm)|
Solubility, transport, and human metabolism of 1 were acceptable, and there was no PGP efflux. A preliminary single PK in rat showed acceptable exposure and bioavailability in spite of a high clearance and short half-life (Table 3). Cytochrome P450 (CYP) inhibition did not appear to be an issue (human recombinant CYP 3A4, 1A2, 2D6, 3C9, 2C19 enzymes, data not shown).
|Sol (mg/mL)||Met Staba,b||Met Staba,c||CaCo2|
|In vivo PKd|
|pH 7.4||pH 2.7||AUC po (h μg/mL)||Fe (%)||Clf (L h/kg)||t1/2 po (h)|
The goals of our optimization program were therefore set by this initial analysis: optimize potency, in particular cellular potency and selectivity versus Syk and other closely related kinases and optimize metabolic stability to achieve a half-life of around 5 h in mice. This article deals with optimization of the potency.
Even though we had no experimental information on the binding mode of 1 in IGF1-R, the SAR data we had supported the hypothesis that the two nitrogen atoms N1 and N7 of the azaindole moiety were interacting with the hinge portion of the kinase (12). Indeed, removing N7 in the bis-indole 2 resulted in total loss of the IGF1-R inhibiting activity. On the other hand, moving this nitrogen atom around the aromatic ring from position 7 to position 4 of the azaindole (data not shown) also led to loss of activity. The fact that methylation on the N1-position of the azaindole showed complete loss of activity, further strengthened this binding hypothesis.
Even though this hit arose from a series that had already been thoroughly investigated in the Syk program, we could gather relatively little information about the SAR for IGF1-R because the 5′-methoxy group was essential to the activity. Table 4 shows the subsequent loss of activity with removal of the methoxy groups on the indole ring. We therefore designed a program of chemical modifications keeping these groups in place.
|Cpd||Position of MeO||IGF1-R (HTRF) (μm)|
Substitution on positions 3 and 2′ on the azaindole and indole respectively also proved to be detrimental to the activity. On the other hand, the 6′-position was more permissive as was also the case for the 1′-position. These positions were heavily explored with production of libraries of various alkoxy, alkyl, and alkyl-substituted side chains, but even if some derivatives were able to, at best, keep the activity of the starting prototype, no substantial gain of affinity was seen (data not shown) and we kept some of these in mind for final modulation of the physicochemical properties like solubility (see later).
Position 4, 5, and 6 on the azaindole were substituted with halogens and a few other groups (Table 5) showing that small hydrophobic substituents on position 4 like chlorine (8), nitrile (9), and alkyls (10 and 12) are well tolerated (IC50’s 43–168 nm) while polar substituents like amine (14, IC50 3 μm) and larger substituents resulted, most of the time, in loss of affinity. Position 5 only accepted limited changes with fluorine (11) even displaying activity enhancement (IC50 19 nm) while activity dropped rapidly with chlorine (13) displaying an IC50 of 400 nm. Disappointingly, the introduction of a primary amine in position 6 as in 16, to improve the affinity by adding a hydrogen bond with the hinge, resulted in loss of activity.
|Cpd||R and position||IGF1-R (HTRF) (μm)|
The crystallization of IGF1-R is complicated by the fact that the receptor exists under various stable conformations depending on its phosphorylation state. At the time we initiated this work, X-Ray data were publicly available for the 0- (13), 2- (14), and 3- (15) phosphorylated forms but we had no in-house experimental co-structure of an azaindole with IGF1-R. A 3D homology model was built using the ATP-bound structure of a 3-phosphorylated form of IGF1-R (PDB code 1K3A). Figure 1A shows a ribbon representation of the ATP-binding site, highlighting the hinge portion which contains the hydrogen bond partners of ATP and the solvent region. In terms of structure-based drug design, our knowledge of this chemical series was enriched by experimental 3D structures solved in-house of various azaindole analogs complexed with Jnk3 and Aurora2 kinase domains. The position of the azaindole scaffold is the same, in all structures, with the azaindole nitrogen atoms hydrogen-bonded to the hinge backbone and the indole substituent lying along the backbone at the entry of the ATP-binding site. Docking calculations allowed proposing two binding modes. The first one, showed in Figure 1B, is similar to the one observed in Jnk3 and Aurora2. The second one places also the indazole moiety in front of the hinge region, but the indole ring is flipped 180° around the C-C bond between the two bicycles, as shown in Figure 1C. This ligand conformation is allowed in IGF1-R because of a small hydrophobic pocket which can accommodate the 5′-methoxy group, with a Leucine residue in the hinge region (Leu1051), where a Tyrosine and a Leucine residue are found in Aurora2 and Jnk3, respectively. Our SAR data in particular in position 5′ along with the selectivity of the compounds versus Aurora2 and Jnk3 supported the hypothesis of the alternative binding mode.
To assess this binding hypothesis, we introduced a nitrogen atom on position 4′ of the indole to stabilize this conformation through an internal hydrogen bond and increase the inhibitory activity of the compound. Quantum mechanics calculations favour this form by 3–4 kcal/mol,1 a value compatible with a hydrogen bond. Indeed, the addition of this nitrogen as in 17 resulted in substantial activity improvement with an IC50 of 6 nm for IGF1-R inhibition.
We finally succeeded in obtaining crystal structures of IGF1-R complexed to azaindazole analogs. Compound 25 was co-crystallized with the bisphosphorylated form of IGF1-R2 demonstrating that the bound conformation is the same as the one observed in Aurora2 and Jnk3, displayed in Figure 1B (Figure 2). Subsequent crystallographic structures with azaindole analogs complexed to IGF1-R showed the same binding conformation (data not shown).
Thus quantum mechanics energy calculations on the unbound 17 did not explain the observed improvement of activity. A hypothesis for the observed activity enhancement of the bis-azaindole series is that the presence of the 4′-nitrogen atom favors the active flat conformation of the two bicycles observed in the crystallographic structure by removing the steric clash between the 3- and 4′-hydrogens in 1. There might even be a small additional stabilization through an internal hydrogen-bond between 4′-N and azaindole 1-H atoms even though this type of H-bond is not very documented in the literature (Figure 3).
We kept this new bis-azaindole scaffold in our final optimization. The best compounds resulting from our chemistry effort are summarized in Table 6. Overall, the 10-fold activity enhancement was observed for all compounds with a nitrogen atom at position 4′ of the indole moiety (17 and 24–28) compared to equivalent analogs with a carbon atom (1,8,11,19–23). This modification has also resulted in a striking impact on the cellular activity, the most potent compound (26) displaying a potency of 1 nm in the MEF IGF-induced proliferation assay. Also, quite interesting was the effect of the aminoethyl side chains at position 1′ of the azaindole on the cellular potency (compare for instance 17 with 24 and 25 with 26). This effect was much less pronounced on the initial carbon series. Compound 26 proved a tight binding IGF1-R inhibitor with a Ki of 1.6 nm (data not shown). It was kept for further profiling.
|Cpd||X||R’1||R4||R5||IGF1-R (HTRF) (nm)||IGF1-induced proliferation (MEF) (nm)|
Compound 26 was profiled in a panel of proliferating cell lines. The huge improvement of potency in the MEF cell line did not translate into the same improvement versus these additional cell lines. The best responding cell line turned out to be HT29, a colon cell line, with an IC50 of 140 nm (Table 7).
|Tissue||Cell line||IC50 (μm)|
The general synthesis of the 2-(3-indolyl)-7-azaindoles 4, 5, 8 11, 13, and 18–22 (compounds I) is outlined in Scheme 1 (16). The substituted 1-Boc-3-boronic acids 5-methoxy-indoles C were obtained from 5-methoxyindole A after iodination in position 3 under basic conditions followed by iodine lithium exchange and trapping with trimethylborate. The substituted 7-azaindoles F were synthesized from the corresponding 2-aminopyridines D after 3-iodination, displacement of the 3-iodine by a trimethylsilylethynyl (17,18) group, and cyclization under basic conditions. The resulting azaindoles F were protected on the nitrogen by a tosyl group and iodinated under basic conditions to afford the 2-iodo derivatives G. These two building blocks C and G underwent palladium-catalyzed coupling to afford the bis-bicyclic derivatives H. Alkylation on the indole under basic conditions followed by deprotection of the tosyl group afforded the final compounds I. From these, functional modifications led to the azaindoles 9, 10, 12, and 141.
The general synthesis of the 2-[3-(4-azaindolyl)]-7-azaindoles 17, 24–28 (compounds R) is outlined in Scheme 2 (19). The 4-azaindole-building block N was obtained via a quite original synthesis: the 3-hydroxy-2-bromo-pyridine J was converted to the 2,3-dimethoxy-6-iodo-pyridine K after iodination (20) in position 6 followed by methylation of the phenol and substitution of the 2-bromine by sodium methoxide. Metal lithium exchange with nBuLi followed by formylation with Dimethylformamide (DMF) afforded the 6-formyl derivative which was nitrated at position 5 with cupric nitrate trihydrate (21). Condensation with nitro methane under classic Henry conditions (22,23) followed by a modified procedure reported by Novellino (24) gave the hydroxylated nitrostyrene which upon heating with sodium acetate in acetic anhydride gave the nitrostyrene intermediate L as a mixture of isomers. Silica gel-assisted reductive cyclization (25) of this unstable intermediate, with iron in acetic acid, provided the desired 4-azaindole M. This compound was further brominated in position 3 and protected with Boc-carbonate to afford N. On the other hand, the 7-azaindoles O were protected with a tosyl group which then allowed the specific introduction of the tributyltin moiety in position 2 to afford P. Compounds P and N were coupled using standard Stille conditions and after subsequent deprotection of the Boc and tosyl groups, followed by N-alkylations the desired compounds R (cpds 17, 24–28) were obtained (Scheme 2).
A lead from HTS was optimized for the inhibition of IGF1-R using rational drug design. This led to the discovery of a very potent series of bis-azaindoles, although the subsequent crystallographic experimental data did not confirm the binding mode proposed by the in silico calculations.