Correction : Synthesis of kinase inhibitors containing a penta fl uorosulfanyl moiety †

The incorrect CCDC numbers (154150–154153) were quoted in the footnote referencing the electronic supplementary information (ESI) and also in the Experimental section. The correct ESI footnote is shown below with CCDC numbers 1564150–1564153. The CIF files posted originally were also incorrect. These were replaced with the correct files on 29th November 2017. The Royal Society of Chemistry apologises for these errors and any consequent inconvenience to authors and readers.


Introduction
The dysregulation of protein phosphorylation mediated by protein kinases is key to the progression of a number of cancers. Unsurprisingly, a number of ATP-competitive kinase inhibitors are in clinical use and development. [1][2][3][4][5][6][7] For example, the oxindole-containing antiangiogenic drug Sunitinib 1, containing a 5-fluorine substituent and a solubilizing side chain on the pyrrole unit, is in clinical use and superseded Semaxanib (2, SU5416) ( Fig. 1) as well as inspiring a number of other studies on druglike oxindoles. [8][9][10][11][12][13][14][15] Metal-based analogues such as 3, 4 have been described by our group and show kinase inhibition down to the nM range and tolerance of a range of substituents at the C-5 position. 16, 17 Meggers's group replaced the sugar unit in staurosporine, a pan-kinase inhibitor with relatively high toxicity and unsuitable for clinical use, by square planar and octahedral transition metal complexes 5-7, leading to highly potent, selective kinase inhibitors. This was attributed to the novel "imaginary hypervalent carbon" geometry enabled by the metal complexes (Fig. 2, 5-7). [18][19][20][21] The pentafluorosulfanyl group is attracting increasing interest in medicinal chemistry. Displaying strong polarity, high lipophilicity and good stability under physiological conditions, an SF 5 substituent has often been shown to behave like a CF 3 group. [22][23][24][25][26] Here we show that a SF 5 group can be incorporated in both classical and metal-based oxindole derivatives, at the 5-or 6-position, leading to analogues displaying kinase inhibition down to the nM range.

Results and discussion
Microwave-mediated Knoevenagel condensations of the commercially-available 5-or 6-SF 5 -substituted oxindoles 8 27 with three separate aldehydes led to the products 10-14 (Scheme 1). 28 The structures of the pyrrole-containing positional isomers 10 and 11 were confirmed by 1 H NMR, 13 C NMR spectroscopy, elemental analysis and mass spectrometry. In their 1 H NMR spectra the most downfield signals were assigned to the pyrrole-NH groups (δ 11.10-13.40 ppm) due to an intra- molecular NH⋯OvC hydrogen bond and further confirmation of their anticipated Z-configuration and such a hydrogen bond was provided in the solid state (Fig. 3). 29 The related reaction with ferrocene carboxaldehyde afforded a mixture of stereoisomers 12a and 12b, which were separated by chromatography. Both isomers were characterized in the solid state (Fig. 4).
We tested all synthetic compounds against a panel of kinases in a biochemical assay. Each data point was measured in duplicate (technical replicates). The potencies of compounds that showed appreciable (approx. 50%) inhibition at 1 μM concentration were established by testing them over a dose range to determine their IC 50 values. Additional kinase binding studies were performed vs. a select group of functionally and structurally divergent kinases including AAK1 (Adaptor-associated protein kinase 1), BMP2K (BMP-2-inducible protein kinase, where BMP is bone morphogenic protein), GAK (Cyclin G-associated kinase) and STK16 (Serine/threonine-protein kinase 16) ( Table 1). In all assays a control of staurosporine, a known promiscuous kinase inhibitor, was used.
In the case of a number of kinases, e.g. VEGFR2 (vascular endothelial growth factor receptor 2) and DYRK2 (Dual-specificity tyrosine phosphorylation-regulated kinase 2), no appreciable inhibition was observed for any of our synthesized compounds, suggesting that we might observe differences in their selectivity, i.e. no promiscuity, towards this panel of kinases. Compound 10 bound to BMP2K with an IC 50 of 452 nM whereas 11 displayed nM potency vs. PDGFR2 (98 nM) and submicromolar potency vs. VEGFR3 (230 nM). Stereoisomeric 12a and 12b only inhibited DYRK3 in the low micromolar range. The positional isomers 13 and 14 both inhibited VEGFR3 with IC 50 s of 530 and 18 nM respectively whereas the latter displayed an excellent 3.1 nM IC 50 vs. PDGFRα.
The synthesized compounds were next tested in breast cancer and non-transformed breast cell lines. Compounds 10 and 11 potently inhibited MCF7 and T47D breast cancer cell proliferation with GC 50 values ranging from 0.35 to 3.8 μM with compound 11 proving superior to compound 10.
MCF7 and T47D cells are luminal A ER + /PR + /HER2 − cells that would normally be responsive to estrogen and progesterone receptor (ER/PR) antagonists such as tamoxifen and megestrol respectively, but not to human epidermal growth factor receptor 2 (HER2) inhibitors. MDA-MB-231 (abbreviated as MM231) cells are triple negative (ER − /PR − /HER2 − ) and cannot be treated with hormone receptor and EGFR (HER2) inhibitors, making cancer cells such as these refractory to most treatment strategies. Compounds 10 and 11 may offer advantages for the treatment of ER + /PR + cancer cells by poly-   pharmacologically targeting multiple kinases such as the receptor tyrosine kinases and other serine/threonine kinases. Lastly, it is encouraging that normal MCF10A cells were resistant to all inhibitor treatments suggesting these compounds would have a large therapeutic window ( Table 2). Compound 11, which bears a methylidene indolinone scaffold ( Fig. 1), demonstrated its greatest potency against the receptor tyrosine kinase PDGFRα, which adopts an inactive conformation according to X-ray crystallographic analysis ( Fig. S1B †); however, an X-ray co-crystal structure containing a methylidene indolinone-based inhibitor (15, Fig S1 †) bound to the RET kinase domain reveals a type 1 inhibitor bindingmode, or binding to an active kinase conformation (Fig. S1B †). Alignment of 15-bound RET with the PDGFRα structure reveals gross structural shifts between analogous β-hairpins and Cα-helices, which is not surprising as the active conformation is generally rigid and condensed and the inactive conformation is generally more open. 30 Alignment of the Dasatinibbound co-crystal structure of Protein-tyrosine kinase 6 (PTK6), a non-receptor tyrosine kinase, with the 15-bound RET reveals that they share a similar, active conformation ( Fig. S1C †). Based on this analysis, it makes sense to use an active kinase conformation, as the above elements (β-hairpin and Cα-helix) are proximal to the ATP-binding pocket and likely to have an impact on binding mode. However, rather than performing docking studies with RET, we decided that PTK6 would be superior as this kinase has a threonine gatekeeper residue, similar to that of PDGFRα, whereas RET has a valine at the same position. Valine is slightly bigger and more hydrophobic than threonine, lacking a hydroxyl group compared to threo-nine, and could drastically perturb interactions necessary for 10 and 11-binding. Furthermore, based on the similarity of 10 and 11 with other type 1 methylidene indolinone inhibitors, we predicted that docking these compounds to an active PTK6 kinase conformation would yield improved binding energies; a result confirmed by docking 10 and 11 to the inactive kinase conformation of PDGFRα (PDB: 5K5X), which reported higher binding energies, and thus less avid binding, for both 10 and 11.
Against PTK6, both compounds bind in a very similar manner as seen in Fig. 5 (top panel). We found the SF 5 moiety of 10 and 11 to bind deeply in a predominantly hydrophobic   pocket next to the gatekeeper residue (Fig. 5 top and bottom  panels). The amide hydrogen of both compounds interacts with the Met267 backbone; however, note that the attachment of the SF 5 group to position 5 of the oxindole ring forces compound 10 to swing away slightly from the hinge. This may explain why inhibitor 11 is more potent in cells and in vitro (PDGFRα & VEGFR3) as the hydrogen bond distance is shorter for the 11 docking-pose, indicative of a stronger interaction.

Conclusion
A small library of SF 5 -containing oxindole analogues has been synthesized. Many products were characterized in the solid state and assayed vs. a small panel of kinases. Docking studies predicted effective binding of the SF 5 group to a hydrophobic cleft in the kinase and biochemical assays showed little evidence of promiscuity in the range of analogues synthesized. This bodes well for the use of the SF 5 group in medicinal chemistry with compound 14 in particular showing low nM potency against VEGFR3 and PDGFRα kinases.

Conflicts of interest
There are no conflicts to declare.