Open science experiment #2: NO ANIMALS PER RECENT NIH GUIDANCE

Operational model R.I.P.?

Goal: Create a non-psychedelic 5-HT2A agonist with therapeutic utility without relying on animal models.

Reminder: we are using classical in vitro pharmacology (NO ANIMALS PER NIH GUIDANCE) for all of these studies I will present here.

Last post on open science I showed these sorts of ‘cruelty-free’ plots (all generated with recombinant receptor expressed in a tet-inducible manner in HEK293 cells):

The upper plot is calculated using the ‘peak height’ method while the lower plot is calculated using the ‘AUC’ (area under the curve) which gives a more integrated response. If we choose either lisuride or Br-LSD as our ‘non-psychedelic’ drug control, our data show modest differences in the apparent potency (e.g. EC50) but no real difference in efficacy (e.g. Emax) compared with the classical psychedelic psilocin.
To analyze the data we used a standard 4-parameter dose-response equation:

The analysis revealed the following parameter estimates for the “peak height” below:

The analysis revealed the following parameter estimates for the ‘AUC’ approach:

The ‘Span’ is equivalent to Emax. As can be seen there are potentially significant differences in the various parameters depending on what method was used to calculate the response.

We have arbitrarily chosen Br-LSD as our exemplar non-psychedelic neuroplastogen and, according to Google we are not alone in our choice:

“…BetterLife Pharma is developing 2-Br-LSD for cluster headaches and other indications, while Seaport Therapeutics is developing a prodrug of 2-Br-LSD (SPT-348) for depression, anxiety, and other neuropsychiatric disorders.”

So the first question we asked is: is there any way to analyze the data to provide an estimate of the ‘intrinsic activity’ vs. efficacy of a compound. As can be seen in the plot below, Br-LSD’s actions are highly dependent upon receptor expression:

ditto for psilocin:

Here is the comparison at 2 and 4 hr (low-ish levels of expression; similar to brain):

Our first effort at analyzing the data was to use the Operational Model originally formulated by Black and Leff (Sir James Black cites this in his Nobel lecture). An important paper published by Terry Kenakin (the expert on quantitative pharmacology and a colleague at UNC) and others provides this extension of the operational model

and the equation that can be used to fit such data is shown here:

The paper (and method along with underlying assumptions) predicts that a parameter referred to as the

Δlog(τ/KA) value

will be uniform for all tested agonists and partial agonists. This value is obtained by calculating the Log(τ/KA) (transduction coefficient) for each test agonist and then to determine the Δlog(τ/KA) by determining the difference between the Log(τ/KA) for the reference full agonist and the test agonist.

As shown in this paper with some nice examples this value Δlog(τ/KA) should be invariant as shown in this classic study of ‘spare receptors’ eliminated with pre-incubation with phenoxybenzamine (POB) (e.g. high—>low receptor expression):

Soooo….we (XP) calculated the Δlog(τ/KA) and the results are shown below (AUC):

Sadly, the prediction of the operational model was not confirmed.

We are going to repeat this experiment (which I’ll show next post) to determine if these results are reproducible with some more interesting ligands and several distinct transducers. Assuming we can reproduce these results with more ligands and different transducers I’ll go on to explore what the next step would be.

For now, it would appear that using this simple read-out for our studies is going to be problematic.