Here’s an unusual paper that’s studying receptor behavior on cell surfaces by use of atomic force microscopy. (Here’s the SI file, which is free to access). The authors took the marketed VEGF inhibitor vandetanib (VD6474) and attached it through linkers to the AFM tip, and then scanned around the surface of live human umbilical vein endothelial cells (HUVECs) sitting on a test surface. Using a tip that was basically saturated with ligand, they were able to detect binding as they got closer to the cell surface, and could map where the effect was strongest by repeatedly pulling the probe back up and repositioning it.
That indicated the strongest binding around the periphery of the cells, which was in line with immunostaining results on the locations of the VEGF receptors. Switching to an AFM tip with much lower compound loading and running a similar set of experiments allowed the team to estimate the force involved in pulling a single vandetanib molecule back out of its VEGF binding pocket (about 45 piconewtons, if you’re wondering). They also estimated that there were about 16 receptors within range of the probe tip, on average (an area of about 250 square nanometers), and the maximum separation between VEGF receptors in that region was about 48 angstroms.
With those figures in mind, the group set about making some multivalent forms of vandetanib, shown at right with the drug structure itself in orange. These had various geometries and spacings between the ligands, and each one was studied in a receptor binding assay. The parent compound has an IC50 of about 47 nM in the assay, for reference, and the bivalent one to its right improves to 1.8 nM. The next one (ZD-3) is identical, at 2.2 nM, and the one at the bottom is only a bit better, at 0.9 nM. These improvements are probably just due to the statistical increase in the likelihood of binding, as expected with a polyvalent ligand, but the ZD-4 structure in the middle seems to hit the spacing just right. That one stands out at 0.025 nM (25 picomolar), a 100-fold improvement over what just multivalency seems to provide and a 2000-fold improvement over the patent drug.
Interestingly, the paper goes on to try that compound out in vivo, comparing radiolabeled versions of the parent compound and the tetravalent form in a U87 xenograft mouse model, dosed i.v. Imaging and histology showed that the tumor uptake of the latter was at least 12-fold higher than the parent drug, and considering the pharmacokinetic challenges of a species like that one, that’s pretty good. There are two comparison PET images in the paper at 24-hour time points, and if I wanted to cavalierly extrapolate from them, I’d say that the parent drug’s radiolabel is showing up more in the urine by then compared to the tetravalent dosing, and the latter seems to be piled up more in the liver (as well as definitely showing higher tumor concentration). But that’s speculation – it would be worthwhile to see a full PK workup on the ZD-4 species itself.
There are a lot of things to be tried out, given these results: how much, for example, does the spacing and density of VEGF receptors vary between cell types and under different conditions? How dynamic is that distribution for even a single cell? What happens if you apply this technique to other cell-surface receptors – how general is it? And there’s surely a lot of variation to be tried on the multivalent chemical structures themselves. One thing that’s clear from the history of such hybrid molecules is that the length and chemical character of the spacing elements is often important, and in ways that are not always easy to rationalize. I’m tempted to say that the team here got very lucky if they did only prepare the four varieties shown in the paper, but there’s nothing wrong with luck, either, is there?