Selective Raf inhibition in cancer therapy

Abstract

Over the past 5 years, the Raf kinase family has emerged as a promising target for protein-directed cancer therapy development. The goal of this review is to first provide a concise summary of the data validating Raf proteins as high-interest therapeutic targets. The authors then outline the mode of action of Raf kinases, emphasizing how Raf activities and protein interactions suggest specific approaches to inhibiting Raf. The authors then summarize the set of drugs, antisense reagents and antibodies available or in development for therapeutically targeting Raf or Raf-related proteins, as well as existing strategies combining these and other therapeutic agents. Finally, the authors discuss recent results from systems biology analyses that have the potential to increasingly guide the intelligent selection of combination therapies involving Raf-targeting agents and other therapeutics.

Figure 1. The core Raf signaling pathway

Components of the central activation cascade proceeding from ligand-bound EGFR through Raf to the nucleus are indicated in green or yellow boxes, connected by red arrows. Additional proteins regulating Raf are indicated in gray boxes. Additional Raf phosphorylated/regulated proteins beyond the central cascade are indicated in pink boxes: not all are shown (see text). After Ras stimulation, B-Raf heterodimerizes with c-Raf-1; mutationally activated B-Raf binds and activates C-Raf and MEK independent of Ras [170]. Therapeutics targeting B-Raf and c-Raf1 are the predominant topic of this review: agents targeting EGFR and MEK1 are also briefly discussed.

Figure 2. Domain structure and key phosphorylation sites for B-Raf and c-Raf-1

This is a simplified depiction of the regulatory phosphorylation sites governing activity of the Raf kinases. Darker red region within the kinase domain indicates the catalytic loop. Key activating phosphorylations on c-Raf-1 (shown in red) include S338, Y341, T491 and S494. Akt confers an inhibitory phosphorylation on S259, which is dephosphorylated during c-Raf-1 activation. Note that on B-Raf that S446 (equivalent to S338 of C-Raf) is constitutively phosphorylated, while D448 (positionally equivalent to Y341 on c-Raf-1) is a phospho-mimic. These differences are thought to contribute to the greater ease of mutationally activating B-Raf1 through a single V600E mutation. Because of its relatively minimal contribution to cancer, A-Raf is not shown; the structure and regulation of A-Raf are similar to those of c-Raf. Darker red region within the kinase domain indicates the catalytic loop. Additional phosphorylation sites indicated represent basal/constitutive phosphorylation sites that contribute to Raf interactions. See detailed discussions of regulatory phosphorylation of Raf in [32, 165, 169]

Figure 3. Comparison of inactive and active Raf kinase

Inactive Raf is locked in a closed conformation anchored by binding of 14-3-3 to N- and C-terminal phosphorylated residues on Raf. Sequential dephosphorylations by PP2A, followed by activating phosphorylations, lead to opening of the Raf conformation, interaction with Ras and recruitment to the membrane, and ability to bind and phosphorylate substrates such as MEK1. Additional Raf-interacting proteins that support these processes (e.g., the chaperone HSP90; the scaffold KSR) are not shown.

Figure 4. Chemical structures of Raf-targeting agents

See text and Table 3 for additional details. Structures for AAL881 and LBT613 are currently unavailable. Structure provided for Raf265 represents patent example.

Figure 5. The Raf interaction neighborhood

The program STRING (‘Search Tool for the Retrieval of Interacting Genes/Proteins’, http://string.embl.de/) collects known physical interactions from the following databases: BIND, DIP, MINT, BioGRID, and HPRD. Each interaction is annotated by STRING with a benchmarked numerical confidence score. In addition, STRING can retrieve indirect protein associations (e.g., genetic and functional interactions) from pathway databases (e.g., http://pid.nci.nih.gov), and by text mining the scientific literature. C-Raf-1, A-Raf, and B-Raf interactions were collected with a cut-off score of 0.4 (medium confidence) for experimental data only; 0.9 (highest confidence) for text mining; and 0.98 for pathway mining. Data were imported in the Cytoscape software (www.cytoscape.org). In addition, BIND, DIP, MINT, BioGRID, HPRD and Intact databases were also searched using the BioNetBuilder plugin for Cytoscape and additional search tools. All data were imported in Cytoscape and merged. Although there are numerous interactions among the group of Raf family-interacting proteins, only direct interactions with Raf proteins are shown here (a full version is available on request). Nodes (circles; indicating discrete proteins) were color-coded according to confidence level of the interaction with Raf as follows: pink, >0.9; green, >0.7; blue, > 0.4; yellow, not rated (e.g. the database providing the information lacked sufficient annotation for assignment). In addition, the edges (lines) are color-coded as follows: blue, direct protein-protein interaction; red, pathway maps; green, text mining.