. I suppose that it could be possible for a drug to bind to a conformational form which supports G-protein activation, but not downreuglation... but I doubt it.
Check out the latest findings about Gi vs. Go signal transduction in morphine tolerance (which has been led to by studying the effects of ultra-low-dose naloxone). Pretty interesting. I'm not an expert on GPCR, but I think it might be possible that different opioids modulate different types of signalling by interacting, or not interacting, with specific amino acids of the protein. There's too much we don't know or understand about opioid tolerance to cast so much doubt about finding an opioid with reduced dependence potential - but obviously it should be studied more.
I just plain don't trust most animal models. Animal models for stroke, depression, MS, epilepsy etc are shit.
MS, stroke - definitely. Epilepsy - don'y know. Depression? Animal models to date have been pretty good predictors for efficacy in humans.
I think opioid testing in animals has also proven pretty reliable, though different species, and even different strains of the same species, have been found to have quite different reactions to the same substances.
If you can show me that this drug is a selective mu agonist, doesn't cause mu receptor internalization in primary neuronal cultures and doesn't cause a reduced transduction efficacy in regards to a fast correlate of receptor activation
Quite the contrary, as much as animal models can be problematic, actually showing reduced dependency in a living being is far better support than any suspected correlate of in vitro biochemistry.
Also, the role of MOR internalization is unclear. Morphine, for example, causes mininmal receptor internalization and desensitization. (some researchers believe receptor internalization is crucial for preventing tolerance).