Neurotrophic Factor Mimetics: Target Validation to Lead Optimization

(Frank M. Longo, Stanford University)

Case study: small molecule ligands for neurotrophic receptors

  • Research started at UCSF, continuing at Stanford; current virtual company (Pharmatrophix) for commercial development

Research started at UCSF, continuing at Stanford; current virtual company (Pharmatrophix) for commercial development

Key decision point #1 – Biological validity of target

  • Try to use as unbiased an approach as possible to consider the biological robustness of the target
  • Consider which disease is best suited – oftentimes a single target may be relevant for multiple diseases
  • Example: p75NTR as a robust target for AD, with many signaling pathways tied in with those known to be affected by pathogenic levels of Abeta (see REF – Knowles)
    • Mechanistic approach – specifically target the known pathways that lead to a beneficial response
    • For p75NTR, modulate the receptor to recruit “survival” adapters and stimulate PI3K activity, which can counteract some of the pathogenic downstream effects of Abeta exposure
  • Use mouse models, if available, to test target validity with in vitro and in vivo experiments.
    • Neuron cultures from the p75NTR knockout mouse did not exhibit neuritic dystrophy when exposed to high levels of Abeta.
    • hAPP transgenic mice crossed with the p75NTR knockout mice resulted in significantly decreased neuritic dystrophy around Abeta plaques.

Key decision point #2 – Lead identification of small molecule ligands

  • Ligand-focused virtual screening (in silico) can enhance lead identification.
    • Previous p75NTR HTS screens failed so a virtual screening approach was taken to identify small molecule hits by focusing on ligand structure and binding (see REF – Massa).
  • Hits from virtual screens can be tested with in vitro assays and then cycled back through the virtual screening process to find more structurally-related hits. Medicinal chemistry is used to further optimize lead compounds.
  • With this process, you still need to prove that identified hits are indeed ligands of the target.
    • A priori virtual screening could bias hits towards desired target ligands but not necessarily.
    • Need to use assays to demonstrate binding and ability to affect target pathways.
    • Multiple p75NTR assays were done to demonstrate receptor binding and survival-promoting activity.
  • Keep in mind that while the identified hits, as with most drugs, may also have off-target effects, these are nearly impossible to disprove.

Key decision point #3 – Disease relevance

  • Once hits are obtained, establish disease relevance using in vitro and/or in vivo models.
    • The identified p75NTR hits are able to ameliorate several Abeta-induced pathological effects: neuritic dystrophy, dysregulated signaling pathways, and synaptic plasticity deficits (see REF – Yang).
  • A small molecule ligand for a receptor may actually have entirely different effects from the natural ligand.
    • In a cell death assay, a p75NTR small molecule hit prevented neuronal apoptosis induced by Abeta exposure. Interestingly, the natural ligand, NGF, did not have this beneficial effect.

Key decision point #4 – Compound selection for in vivo studies

  • Factors to consider in selection:
    • Predictive algorithms performed by a medicinal chemist (e.g. Lipinski’s rule); use software prediction when possible
    • BBB penetration, PK assays
    • Toxicity from informal acute/chronic dosing

Key decision point #5 – In vivo disease models

  • There are pros and cons for the multiple types of disease models. Factors to consider:
    • Relevance
    • Mechanism-relevant endpoints
    • Biomarker-relevant endpoints (important for translatability)
    • Types of behavioral studies
  • AD mouse models have a poor track record for predicting translatability into the human disease (see REFS – ADDF review, Zahs, Ashe)
    • Available AD mouse models are generally agreed to be more representative of the pre-clinical, pre-MCI disease state (and therefore may not be relevant for therapeutics targeted towards clinical MCI/AD).
    • Additionally, there are few endpoints which are consistently reported using these mouse models – the majority of studies include Abeta levels/accumulation but very few measure neuritic changes (see FIG. 1).
  • Improve the robustness of the data by using multiple mouse models at more than one age (including aged mouse models with well-established pathology).
    • Also measure multiple neuronal and non-neuronal endpoints, including more than one behavioral task, to confirm therapeutic effects.
    • p75NTR ligand improves various measures of pathology in multiple AD mouse models (e.g. spine density, cholinergic innervation, glial activation, behavioral tasks).
  • Results should be repeated by other research groups
  • Consider multiple biomarker-relevant endpoints that are potentially translatable for use in a clinical trial.
    • Possible biomarker assays for p75NTR ligand treatment:
      • FDG-PET glucose imaging – detect cholinergic innervation (See REFS – Mega, Dubois)
      • MG-PET (PK1195) microglial activation imaging (See REFS – Venneti, Cagnin)
      • CSF tau level

Key decision point #6 – Decrease risk

  • Test non-specific toxicity, mechanism-based toxicity, and human pharmacology.
    • Early removal of risky compounds with unfavorable hERG/CYP450/Ames results

Key decision point #7 – Academic vs. commercialization

Key decision point #8 – Commercialization pathways

  • University license to existing company
  • Start and fund new company
Academic Models of Drug Discovery: Services and Utilizing CROs >