Frank-Kamenetsky et al.
] chose to use chemical genetics
in an attempt to identify compounds that could interfere with the inhibition of Smo by Ptc, or could activate Smo independent of Ptc, or that might act downstream of Smo. In addition to making effective drug candidates, these molecules might also help to illuminate the complex mechanisms underlying signaling by Hh, Ptc, and Smo (see the box for more of the background to the work).
The high-throughput screen was elegantly simple. First, Frank-Kamenetsky et al.
identified a cell line that responds well to Hh stimulation. The introduction of a reporter gene (encoding the firefly luciferase protein) that was turned on by Hh signaling permitted screening for compounds that block or induce signaling by monitoring the expression of the luciferase protein (using a simple luminescence test). They had previously used such a screen to isolate antagonists of Hh signaling, and had demonstrated the effectiveness of some antagonists as potential anti-tumor drugs to treat BCC [7
]. Their new screen of 140,000 synthetic compounds led to the discovery of a few candidate agonists that could stimulate the reporter gene and mimic Hh activity.
Once the cell-based screen was completed, the chemists took over, synthesizing 300 derivative molecules until they found a few compounds that were related to the previous 'best' agonist but that were a thousand times more effective at eliciting a cellular response, affecting cells when applied in the nanomolar range. "Getting more potent compounds was essential if we were to figure out where the agonists were acting" recalls Jefferey Porter who headed the team at Curis, Inc.
Then the cell biologists began again, studying the effects of the agonists on the proliferation of primary neonatal cerebellar granule neuron precursors. They monitored the incorporation of tritiated thymidine into cultured rat neurons (as a marker of DNA synthesis and hence proliferation) and were pleased to see that the agonists were as effective in this assay as the Hh protein itself. An assay using an explant of embryonic neural plate was used to confirm that the agonists could induce dose-dependent gene expression in neural precursors, just as the Shh protein does.
Having established the effects of the agonists in culture assays, the researchers then turned to an in vivo model, feeding the compounds to pregnant mice and following the effects on the phenotypes of embryos lacking Shh or Smo. The treated embryos displayed activated Hh signaling, demonstrating that the compounds were not toxic and could cross the placental barrier. The developmental defects of Shh-/- embryos were rescued by treatment with the agonist but the compound had no effect on Hh signaling in the absence of Smo.
"Once we knew that the agonists were targeting Smo, we wanted to investigate whether they bound to Smo directly and how they activated Hh signaling," says Porter. The cell line that was created for the screen served as a useful tool to test the effects of known antagonists on the function of the agonist. An anti-Hh blocking antibody had no effect, so the agonist must work downstream of the Hh-Ptc interaction. But the agonist was blocked by antagonists that work at the level of Smo or further downstream. "We have similar conclusions," says Beachy, whose group used photo-affinity labeling and cross-linking experiments to show that small-molecule agonists and antagonists bind directly to Smo.
Next, for Frank-Kamenetsky et al., it was time for some careful biochemistry. Analysis of the expression of fusion proteins of Ptc or Smo showed that, unlike the Hh ligand itself, the agonist had no effect on the stability of the Ptc protein. In contrast, both Hh and the agonist could increase the stability of the Smo receptor. Immunoprecipitation experiments with radiolabeled agonist showed that the agonist must bind directly to Smo receptors, and that Hh-signaling inhibitors compete with the agonist for binding. Pharmokinetic analysis provided evidence for a single binding site competition model. Finally, Frank-Kamenetsky et al. exploited an oncogenic, constitutively active mutant form of Smo (Smoact); the agonists bound equally well to mutant and wild-type forms, whereas the antagonist bound less well to the mutant form.