THE NEUROPLATFORM : TOWARDS A COMPREHENSIVE MECHANISTIC QUANTITATIVE SYSTEMS PHARMACOLOGY MODEL OF ALZHEIMER’S DISEASE FOR SUPPORTING DRUG DISCOVERY AND DEVELOPMENT

Hugo Geerts

Certara

​

Alzheimer’s Disease : The Problem

Treatment Options

-tacrine (1992)

-donepezil (2000)

-galantamine* (2000)

-rivastigmine (2000)

-memantine (2004)

…

>200 trial failures

…

• Aducanumab (2021)
• Leacanemab (2023)
• Donanemab (2024?)

• Over 55 million people with dementia in the world
• Cost 400 billion$/year (US)

Courtesy of R. Tanzi

2

© Copyright 2020 Certara, L.P. All rights reserved.

From Molecular Biomarkers to Clinical Function in AD

3

Clinical

Scales

Synapse

EEG,

BOLDfMRI

Network

Gene-gene

coregulated

Causal networks

ADAS-Cog

CDR-SOB

Clinical Biomarkers

Functional Biomarkers

Neuro-Fibrillary Tangles

Amyloid plaques

Microglia

3

© Copyright 2020 Certara, L.P. All rights reserved.

Outline

• Computational Neuropharmacology
• Can we simulate functional clinical scales?
• Amyloid Aggregation Modeling
• Modeling dynamics of “invisible” intermediate Abeta species to predict biomarkers and functional outcome
• Tau Pathology Modeling
• Addressing the challenges of Tau Therapeutics
• Virtual Twin
• Reducing clinical outcome variability

​

​

4

© Copyright 2020 Certara, L.P. All rights reserved.

Computational Neuroscience:�How can we maximize drug effects on symptoms? ��

Simulating Neuronal Network Activity as Proxy for Functional Outcomes

5

Simulating Neuronal Circuits in a Biophysically Realistic Way

Brain regions are represented by synaptically coupled biophysical models of neurons

Both imaging studies (BOLDfMRI) and icEEG) and interventional studies (rTMS) suggest neuronal firing drives domain-specific clinical outcomes.

In Silico Biosciences

Applying Hodgkin-Huxley equations to predict firing dynamics in specific neuronal networks that drive clinical outcomes

6

© Copyright 2020 Certara, L.P. All rights reserved.

INTERMEZZO 1 : NEURON MODEL EQUATION

7

© Copyright 2020 Certara, L.P. All rights reserved.

Raster diagram readout

A) Model Neurons

B) Microcircuits

C) Full Circuit

The nuclei (thalamus, cortex, basal ganglia) are synaptically interconnected with known anatomy to form the full model.

Synaptic strengths are adjusted so that the spontaneous spiking activity is consistent with electrophysiological data in the literature.

This state of the model represents the “healthy” case.

8

© Copyright 2020 Certara, L.P. All rights reserved.

Introducing drug exposure and target engagement

A) Model Neurons

B) Microcircuits

C) Full Circuit

D) Drug Pharmacology & Target Engagement

The effects of neuromodulators (dopamine, serotonin, etc) are added to the model by computing the activation of receptors with a receptor competition model.

E) Physiological coupling

9

© Copyright 2020 Certara, L.P. All rights reserved.

Defining Pathology and Circuit Emergent Properties

A) Model Neurons

B) Microcircuits

C) Full Circuit

F) Pathology

E) Receptor activation

G) In Silico Biomarker

AD Disease Pathology is introduced using imaging and postmortem data. (1) APOE genotype dependent Synapse loss, (2) Neuronal cell loss and (3) cholinergic state changes

Working Memory span depends on pathology and pharmacological intervention.

D) Drug Pharmacology & Target Engagement

Amyloid pathology affects NMDA, AMPA and GABA leading to hyperactivity

Tau pathology affects Na+ and K+ channels

10

© Copyright 2020 Certara, L.P. All rights reserved.

Calibration with Clinical Data

A) Model Neurons

B) Microcircuits

C) Full Circuit

F) Pathology

E) Receptor activation

G) In Silico Biomarker

Calibrate using published clinical trials

Adjust coupling parameters on receptor effects to best match linear causal relationship with simulated historical clinical trials.

D) Drug Pharmacology & Target Engagement

11

© Copyright 2020 Certara, L.P. All rights reserved.

Amyloid Aggregation Model��

Simulating Effect of anti-amyloid antibodies on biomarkers and cognition

12

Mechanistic Physiology-Based PharmacoKinetic Model

Chang et al. 2019. J.Pharmacokinet. Pharmacodyn.

Amyloid Aggregation QSP Platform

APOE

synthesis

Shorter Aβ40

Longer Aβ42

Based on microscopic kinetic equations (Knowles group -Oxford)

SUVR = Standard Uptake Value Ratio

CSF Ab42

Plasma Ab42

Plasma Ab40

Drug Exposure from PBPK

Geerts CPT : PSP ;12(4):444-461 (2023)

14

© Copyright 2020 Certara, L.P. All rights reserved.

INTERMEZZO 2 : AMYLOID AGGREGATION MODEL EQUATION

trimer gaining monomer

Loss of monomer from pentamer

tetramer losing monomer to trimer

pentamer gaining monomer

trimer gaining monomer on plaque

Tetramer losing monomer on plaque

tetramer cleared by microglia

a(j) bound by antibody

tetramer moving to perivascular space

v

tetramer

15

© Copyright 2020 Certara, L.P. All rights reserved.

Model Can Reproduce Observed Change in Biomarkers

TRAILBLAZER

GRADUATE

16

© Copyright 2020 Certara, L.P. All rights reserved.

QSP Model Predicts Different Biomarker Profile of Lecanemab

Lecanemab has the biggest effect on protofibrils

Plaques

Leca

Adu

Gan

No real difference with aducanumab & gantenerumab

17

© Copyright 2022 Certara, L.P. All rights reserved.

QSP Model Predicts Differentiation on Side-Effect of ARIA-E

Adu

Leca

Time-to-event modeling

18

© Copyright 2020 Certara, L.P. All rights reserved.

Plasma-tau Change is Driven by Soluble Ab42 species

Plaques

R²=0.05

Protofibrils*

R²=0.62

Oligomers*

R²=0.56

Monomers*

R²=0.39

Aggregation

Multiple Regression

R²=0.69

32 observations

7 antibodies

19

© Copyright 2020 Certara, L.P. All rights reserved.

Addressing Treatment Challenges in Clinical Practice

Biomarker of reaching amyloid negativity

Maintenance Dosing

Restarting after ARIA-E interruption

Therapeutic Index in virtual patient trial

Geerts 2024, in press

20

© Copyright 2020 Certara, L.P. All rights reserved.

Soluble Abeta species affect excitation-inhibition balance in neuronal networks

Geerts et al 2018, Geerts et al 2020

Can explain 3 different clinical observational datasets

Can explain clinical effect of BACE-inhibitors

Generate hypothesis for difference between EMERGE and ENGAGE outcome

Huijbers 2018 : Amyloid deposition is related to BOLD hyperactivity

Oligomeric Abeta increases excitability

Complex monomer Abeta effects

21

© Copyright 2020 Certara, L.P. All rights reserved.

QSP Model of Amyloid-Tau Interaction

Z

v

Amyloid aggregation model

PFC microcolumn

Neuronal Activity

Intrasynaptic pTau217

v

Amyloid PET

Monomer dynamics

Oligomer dynamics

CSF/plasma pTau217& CSF tau dynamics

Firing activity

Control : Increased deposition

Increased Ab42 oligomer, increased firing

Aβ42 oligomers

Amyloid antibody : reduces Ab42 oligomer, reduced firing

v

Geerts 2023

Activity-dependent Tau secretion

22

© Copyright 2020 Certara, L.P. All rights reserved.

CDR-SOB changes and changes in Ab42 species

Plaques*

R²=0.41

Protofibrils

R²=0.39

Oligomers

R²=0.35

Monomers*

R²=0.37

Aggregation

Multiple Regression

R²=0.50

34 observations

7 antibodies

23

© Copyright 2020 Certara, L.P. All rights reserved.

Cognition and Neuronal Firing Move in Opposite Direction

Human Encoding uses information bandwidth (Shannon entropy)

Outcome of Computational Neuroscience Platform

MCI and Early Prodromal AD show hyperactivity

Reduced Firing by Amyloid Antibodies

24

© Copyright 2020 Certara, L.P. All rights reserved.

25

TAU QUANTITATIVE SYSTEMS PHARMACOLOGY MODEL

Tau diffusion, interaction with antibody and uptake

20nm

500 nm

26

© Copyright 2020 Certara, L.P. All rights reserved.

Gosuranemab : No Effect Despite CSF Target Engagement

Geerts 2022

27

© Copyright 2020 Certara, L.P. All rights reserved.

Spatial Monte-Carlo Simulation In Synapse Geometry

P. D’Alcre 2015

28

© Copyright 2020 Certara, L.P. All rights reserved.

Seed Internalised Fraction is Not Affected by Antibodies

Lack of sufficient Target Engagement can explain clinical trial failure of first-generation antibodies

8000mg semorinemab

No “bridges’

Bridges spaced 8nm

Bridges spaced 80nm

29

© Copyright 2020 Certara, L.P. All rights reserved.

QSP Model for IntraNeuronal Tau processing

Uptake

Processes

- axonal transport

- oligomerization

- degradation

- secretion at synapse

30

© Copyright 2020 Certara, L.P. All rights reserved.

Virtual Twin Model�Towards Personalized Treatment

Individual Cognitive Trajectories in AD are Very, Very Noisy

ADNI Database

Variability due to

- comedication

- common genotype variants

- Abeta pathology

- Tau pathology

- Various comorbidities

32

© Copyright 2020 Certara, L.P. All rights reserved.

Comedications and Genotypes Affect Cognitive Trajectory

Progression Rate 7-fold difference

Baseline 14 point difference

On same background of amyloid tau pathology

Geerts 2021

Patients on benzodiazepines, antidepressants, cholinergics, antipsychotics

33

© Copyright 2020 Certara, L.P. All rights reserved.

Misfolded Tau affects K+ and Na+ channels

34

Ken Kosik lab

​

Li Gan

V337M

FTD mutation

V337M Tau mutation hIPSC on Multi-electrode Arrays

Create NEURON Model

Identify changes leading to same phenotype

Tau affects K+ and AIS Na+ channels

34

© Copyright 2020 Certara, L.P. All rights reserved.

Using Virtual Twins for Personalized Medicine

Individual target exposure & engagement

Variables

- Comedications

- Genotypes

- Amyloid load

- Tau load

Clinical Trial

outcome

Virtual patient trial

Identical distribution

“Virtual twins”

QSP Virtual Trial

Outcome

Compare on per-patient basis

Novel treatment

Natural Trajectory

Placebo

35

© Copyright 2020 Certara, L.P. All rights reserved.

Conclusion Neuroplatform Properties

• Simulates experimentally inaccessible intermediate Abeta species with biological function
• Provides biological relevant and quantitative relationship between amyloid dynamics and plasma p-tau changes as easily accessible biomarker
• Incorporates effect of amyloid and tau oligomers, comedications and common genotype variants on neuronal firing and cognitive outcome in calibrated QSP model
• Applications
• Support clinical trial design of amyloid, tau, neuroinflammation and combination therapies
• Estimate size of pharmacodynamic interactions with genotypes and comedications
• Support clinical practice decisions
• Identify optimal combination of targets for precision therapy
• Future Directions
• Understanding why clinical outcome with amyloid antibodies declines with increasing baseline pathology
• Role of tau pathology in amyloid antibody response
• Identification of Responders to biomarker changes

​

​

​

​

​

​

36

© Copyright 2020 Certara, L.P. All rights reserved.

THE NEUROPLATFORM : TOWARDS A COMPREHENSIVE MECHANISTIC QUANTITATIVE SYSTEMS PHARMACOLOGY MODEL OF ALZHEIMER’S DISEASE FOR SUPPORTING DRUG DISCOVERY AND DEVELOPMENT

Mike Walker

Rachel Rose

Silke Bergeler

Shaina Short

Viji Chellia

Doug Chung

​

​

Maryam Khalifa

Athena Grant

Jamie Goff

Ivo Schuttel

Athan Spiros

Patrick Roberts

​