Relevance to Drug Development & Patient Stratification Diego A. - - PowerPoint PPT Presentation

Diego A. Pizzagalli, Ph.D.

Professor of Psychiatry Harvard Medical School McLean Hospital

ISCTM Reward Processing February 21, 2020

The Neurobiology of Anhedonia: Circuitry and Relevance to Drug Development & Patient Stratification

• Grant/Research Support:
• NIH, NARSAD, Dana Foundation
• Speaker’s Bureau:
• None
• Consultant:
• Akili, BlackThorn Therapeutics (licensed Probabilistic Reward Task),

Boehringer Ingelheim, Compass Pathway, Otsuka, Takeda

• Stock Options:
• BlackThorn Therapeutics
• Patents:
• None

Disclosures

Role of Anhedonia in MDD & Antidepressant Response 1) Anhedonia predicts:

• Depression two years later (e.g., Wardenaar et al. 2012);
• Poor outcome (e.g., Spijker et al. 2001; Uher et al. 2012);
• Chronic course over 10 years (Moos & Cronkite 1999).

2) Anhedonia and amotivation are poorly addressed by first- line treatments (Calabrese et al., 2014; Craske et al., 2019). 3) Anhedonia predicts poor response to first-line pharmacological (e.g., SSRI; Vrieze et al., 2013) and psychological

(e.g., CBT; McMakin et al. 2012) treatments as well as TMS (e.g.,

Downar et al., 2014).

Borrowing from the “Traditional” Approach…. 1) Since anhedonia has been associated with:

• Reduced functional, structural, and neurochemical

markers within DA-rich regions along the mesocorticolimbic pathways (e.g., ventral and dorsal striatum) (e.g., Auerbach et al., 2017; Gabbay et al., 2017; Keedwell et al., 2005;

Pecina et al., 2017);

2) Dopamine plays a key role in several reward-related functions (incentive motivation, reinforcement learning) 1) + 2) Patients with anhedonic phenotypes might preferentially benefit from treatments hypothesized to increase DA signaling.

Parsing Reward Processing: From Hedonics to Motivated Behavior

Barch et al., 2015

Barch et al., 2015

Barch et al., 2015

Reward Learning

Reward Learning As a DA-sensitive Phenotype

Probabilistic Reward Task 11.5 vs. 13 mm 0.9 vs. 1.6 sec tone

Athina Markou Andre Der-Avakian

Decreased Dopamine Suppresses Reward Learning in Humans and Rats

Hypothesized mechanism: Presynaptic autoreceptor activation→ ↓DA

[Single 0.1 mg/kg dose] [Single 0.5 mg dose] Response Bias

Psychostimulant Exposure Enhances Reward Learning in Humans and Rats

Hypothesized mechanism: ↑ striatal DA transmission?

[Single 0.5 mg/kg dose] [14 mg path in non-smokers]

Der-Avakian et al. Translational Psychiatry 2013

Rats Humans

Barr et al. Biological Psychiatry 2008

N = 30 (14 mg)

Block 1 Block 2 Block 3

Der-Avakian et al. Translational Psychiatry 2017 Barr et al. Biological Psychiatry 2008

Response Bias

Nicotine Withdrawal Suppresses Reward Learning in Humans and Rats

Hypothesized mechanism: ↓ striatal DA transmission?

Rats Humans

Barr et al. Biological Psychiatry 2008

Pergadia et al. JAMA Psychiatry 2014

Reward Learning is Associated with Frontostriatal and Dopamine Markers

Better reward learning: ↓ DAT availability (i.e., higher DA?)

[11C]Altropane (PET) Worse Better Learning

r(31)= -0.43 p=0.01

Reward Learning is Associated with Frontostriatal and Dopamine Markers

Better reward learning: ↑ resting state FC between accumbens and vmPFC r(31)= 0.69 p<0.001

Interim Summary Reward learning: 1) Is associated with individual differences in frontostriatal and dopamine markers (healthy controls); 2) Is potentiated by pharmacological challenges hypothesized to increase striatal DA transmission (amphetamine, nicotine); 3) Is reduced by challenges hypothesized to decrease striatal DA transmission (single low dose of pramipexole, nicotine withdrawal, chronic social defeat); 4) Is reduced in individuals with MDD, especially with elevated anhedonia or melancholia (not shown; Pizzagalli et al., 2008; Liu et al., 2011;

Vrieze et al., 2013; Fletcher et al., 2015)

STUDY 1

Hypothesis:

Patients with MDD failing to respond to SSRI treatment (sertraline) and characterized by pre-treatment anhedonic behaviors will preferentially benefit from bupropion treatment (norepinephrine-dopamine reuptake inhibitor, NDRI).

STUDY 1: Establishing Moderators and Biosignatures

• f Antidepressant Response in Clinical Care (EMBARC)

MDD Patients Sertraline Placebo Sertraline Bupropion XL Sertraline Placebo

Responders Responders Non- responders Non- responders Wk 0 (completed PRT) Wks 1 - 8 Wks 9 - 16

Does pre-treatment reward learning differentiate between eventual responders and non-responders to sertraline and bupropion in Phase 2?

Yuen Ang

N = 262 N = 127 N = 135 N = 60 N = 52 N = 46 N = 73

STUDY 1: Establishing Moderators and Biosignatures

• f Antidepressant Response in Clinical Care (EMBARC)

ANCOVA (Phase 2):

• Drug (SER, BUP) x Response (yes,

no) x Site (CU, MG, TX, UM) [covariates: age, gender and education].

Results (Phase 2):

• Drug x Response: p<0.05
• Bupropion responders have

significantly greater response bias than non-responders (SERT: ns).

• Phase 2 bupropion responders and

non-responders: no Week 0 or Week 8 HAMD differences

Ang et al., under review

Results (Phase 2):

• Drug x Response: p<0.05

STUDY 1: Establishing Moderators and Biosignatures

• f Antidepressant Response in Clinical Care (EMBARC)

ANCOVA (Phase 1):

• Drug (SER, PLA) x Response

(yes, no) x Site (CU, MG, TX, UM) [covariates: age, gender and education].

Results (Phase 1):

• Drug x Response: p>0.45

Ang et al., under review

STUDY 2

Hypothesis:

Patients with MDD and characterized by pre- treatment anhedonic behaviors will preferentially benefit from pramipexole treatment (D2/3 DA agonist)

STUDY 2: Ventrostriatal Dopamine Release and Reward Motivation in MDD (PI: F. Schneier, Columbia) Study Design:

• 26 medication-naïve MDD patients and 26 controls
• Patients received open-label treatment with pramipexole

(ranging 0.5-2.5 mg/day) for 6 weeks

• Before and after treatment:
• Probabilistic reward task (behavior)
• Ventral striatal reward prediction error signals (fMRI)
• Before treatment: Ventral striatal DA release in response to
• ral dextroamphetamine ([11C]-(+)-PHNO PET)
• A priori outcome measures (administered weekly)
• Depressive symptom severity (HAM-D)
• Anhedonia severity (SHAPS)
• Improvement in global illness severity (CGI-Change Scale)

STUDY 2: I. Significant symptomatic improvement following six weeks of treatment with pramipexole

• 72.7% classified as responders at week 6
• Largest effect sizes for depressive symptoms (HAM-D:

d=2.2; MASQ depressive distress subscale: d=1.4) and anhedonia (MASQ anhedonic depression subscale: d=1.3)

STUDY 2: II. Abnormal reward learning, VS reward PE signaling and VS DA release in MDD at baseline

C C

Alexis Whitton Franklin Schneier

STUDY 2: III. Better (i.e., more normative) reward learning and stronger reward sensitivity predicts lower post-treatment anhedonia

Whitton et al., Brain, 2020 PI: Franklin Schneier, Columbia University

STUDY 2: IV. Stronger (i.e., more normative) VS reward PE signals predict greater improvement in global illness severity

Prediction-error signal extracted from ventral striatum

VS gain PE

A

0.5 1 1.5 2 2.5 3 3.5 1 2 3 4 5 6

Predicted CGI scores Week

High PE Mean Low PE (HC group mean)

Whitton et al., Brain, 2020 PI: Franklin Schneier, Columbia University

X = 12

X=-12

STUDY 2: V. Less (i.e., more normative) VS DA release predicts greater improvement in global illness severity

Percentage change from baseline binding potential relative to nondisplaceable compartment (∆BPND) computed from ventral striatum

X = 12

X=-12

C

VS DA release

0.5 1 1.5 2 2.5 3 3.5 1 2 3 4 5 6

Predicted CGI scores Week

High DA release Mean Low DA release (HC group mean)

Whitton et al., Brain, 2020 PI: Franklin Schneier, Columbia University

STUDY 3: FAST-MAS Study (PI: A. Krystal, Duke/UCSF)

B) Monetary Incentive Delay Task

Reward Cue

(500 ms)

T arget

(150 ms)

Feedback

(1,230 ms) You won $5

Loss Cue

You lost $1

No-incentive Cue

No change

+$

• $

0$

Kappa Opioid Antagonist for Anhedonia?

STUDY 3: FAST-MAS Study (PI: A. Krystal, Duke/UCSF)

Snaith Hamilton Pleasure Scale Intent-to-treat sample: JNJ-67953964 (Kappa Opioid

Antagonist, 10 mg) [N=45]

Placebo [N=44]

Krystal et al., Nature Medicine, in press Baseline-adjusted post-treatment scores: (F(1,86)=3.35; p=0.035; Hedges’ g=0.44

STUDY 3: FAST-MAS Study (PI: A. Krystal, Duke/UCSF)

Primary Measure: Nucleus Accumbens Activation to Reward-predicting Cues Secondary Measures: Self-reported anhedonia (SHAPS) Behavior: Response Bias Baseline-adjusted post-treatment scores: F(1,86)=5.58; p<0.01; Hedges’ g=0.58 Treatment Arm x Time: F(1,52)=4.69, p=0.035 [covariate: baseline SHAPS]

Summary

1) As hypothesized, behavioral and neural markers of DA-rich regions within the brain reward system predicted response to pharmacological treatments with DA effects; 2) The direction of the findings was, however, OPPOSITE: More normative Response Bias, Reward Prediction Error, DA release predicted better response to bupropion and pramipexole. → Is a better functioning brain reward system needed to be able to benefit from DA treatments? 3) Prior precedence? 3 of 4 fMRI studies found that MDD individuals with pre-treatment neural patterns more closely resembling controls’ brain function (e.g., during the MID) had a greater response to Behavioral Activation Treatment (Carl et al., 2016; Crowther et al., 2015; Dichter et

al., 2009).

Summary

4) The NIMH “Fast-Fail” approach:

• Target engagement approach
• Pre-screening of patients (transdiagnostic)
• Probe (e.g., fMRI tasks) selected based on hypothesized

mechanisms of the drug (KOR antagonism → reduction of inhibition on DA neurons → potentiated ventral striatal reactivity to reward)

• If failing, fail fast… (and provide interpretable null findings)

Outstanding Questions/Future Directions

1) Pre-selecting/stratifying/enriching using self-report measures (e.g., SHAPS) likely does not yield neurobiologically homogenous phenotypes → Identify “biotypes” using cluster analyses/machine learning?

• Practical considerations of using behavioral batteries in clinical

trials/clinics? 2) “Capitalization” rather than “compensatory” model of change?

• Strategies for non-responders?

R37 MH068376 R01 MH101521 UH3 MH109334 Dana Foundation NARSAD Distinguished Inv. Award

Michael Treadway Roee Admon

fMRI studies

Rosi Kaiser Andre Der-Avakian Athina Markou

Preclinical studies

Acknowledgments

Georges El Fakhri

PET studies

STUDY 3: FAST-MAS Study (PI: A. Krystal, Duke/UCSF)

Primary Measure: Nucleus Accumbens Activation to Reward-predicting Cues Secondary Measures: Self-reported anhedonia (SHAPS) Behavior: Response Bias Baseline-adjusted post-treatment scores: F(1,86)=5.58; p<0.01; Hedges’ g=0.58 Treatment Arm x Time: F(1,53)=3.44, p=0.030; Hedges’ g=0.49

STUDY 2: III. Better (i.e., more normative) reward learning and stronger reward sensitivity predicts lower post-treatment anhedonia

Whitton et al., Brain, 2020 PI: Franklin Schneier, Columbia University

Frontostriatal Resting-state Functional Connectivity Mediates the Relationship Between Striatal DAT Binding Potential and Individual Differences in Reward Learning Behavior

Kaiser et al., Cerebral Cortex, 2018

Rosi Kaiser

Snaith Hamilton Pleasure Scale I would enjoy my favourite television or radio programme I would enjoy being with family or close friends I would find pleasure in my hobbies and pastimes

10 20 30 40 50

HC SCZ MDD

10 20 30 40 50

HC SCZ MDD SUD

Liu et al., 2012 Franken et al., 2007

Role of Dopamine in Depression and Anhedonia?

Direct Evidence

1) Impairments in behaviors known to rely on DA: reward learning (Pizzagalli et al., 2005), effort-based decision making (Treadway et al., 2012), and reward prediction error coding (Kumar et al., 2008, 2018). 2) Reduced reward-related activation within DA-rich ventral (e.g., nucleus accumbens) and dorsal (caudate, putamen) striatal regions (Keedwell et al., 2005; Pizzagalli et

al., 2009; Kumar et al., 2008, 2018).

3) Rapid increases in depressive symptoms after catecholamine depletion among remitted MDD individuals (Hasler et al., 2008, 2009; Homan et al., 2015). 4) Animal models relevant to depression reliably induce anhedonic phenotypes and dysfunction within the mesolimbic DA system (Cabib and Puglisi-Allegra, 2012).

How?

Rhodes et al., 2005

Presynaptic Autoreceptor (inhibitory) Strategy 2: Autoreceptor Blockage (in MDD) Disinhibition of Presynaptic Neuron (↑DA)

Hypothesis:

Pharmacological manipulations that increase phasic DA signaling will “rescue” blunted striatal responses in MDD

Single, low doses of D2/3 antagonists (e.g., amisulpride)*: In rodents:

→ ↑ DA synthesis and release and have prohedonic effects (Coukell et

al 1996; Papp and Wieronska 2000; Schoemaker et al 1997).

*Mechanisms: Presynaptic autoreceptor blockade

In humans:

→↑ ventral striatal fMRI responses to rewards [Jocham et al., 2011] → in MDD, dysthymia: low doses of amisulpride (50 mg/day) have

antidepressant effects (e.g., Amore et al 2001; Boyer et al 1999; Rocca et al 2002; Smeraldi 1998; Zanardi and Smeraldi 2006).

Hypothesized Mechanisms?

Single low doses of D2/3 antagonists block presynaptic autoreceptors (inhibitory) → Increases of phasic DA

Increased DA striatal concentrations

Schoemaker et al., 1997 Jocham et al. , 2011

Increased ventral striatal responses to rewards (200 mg)

Study Design

N = 89 (all unmedicated): 23 MDD + Amisulpride 23 MDD + Placebo 23 HC + Amisulpride 20 HC + Placebo

Plasma level (μg/L) Time (h)

Amisulpride pharmacokinetics

Peak 1: ~1.5 hrs (fMRI with MID) Peak 2: 3-6 hrs (Reward Learning Task)

Coukell et al., 1996

Admon*, Kaiser* et al.,

• Am. J. Psychiatry, 2017

Low Amisulpride Dose “Rescues” Reward Dysfunction in MDD

*

• 0.6
• 0.2

0.2 0.6 1

* *

• 0.6
• 0.2

0.2 0.6 1 1.4

• 0.6
• 0.2

0.2 0.6 1

* *

Amisulpride (50 mg) Placebo

Striatal Response to Reward Outcome

Activation (β) Activation (β) Activation (β)

MDD HC

Caudate Nacc Putamen

Roee Admon Rosi Kaiser

Low Amisulpride Dose “Rescues” Reward Dysfunction in MDD

Amisulpride (50 mg) Placebo

X = 6

*

0.2 0.4 0.6 0.8

*

Connectivity (β)

MDD HC

Nacc- MidCingulate

Striatal connectivity during reward outcome

t = 8 t = -8

Caudate-ACC connectivity

Low Amisulpride Dose “Rescues” Reward Dysfunction in MDD

Amisulpride (50 mg) Placebo

X = 6

Nacc- MidCingulate

Striatal connectivity during reward outcome

t = 8 t = -8

Reward learning (Choose A) Connectivity (β)

• 1

1 2 0.4 0.6 0.8 1

r = 0.65 r = -0.24

For MDD

• nly

Admon*, Kaiser* et al.,

• Am. J. Psychiatry, 2017

STUDY 2

Question:

If MDD is associated with blunted DA striatal signaling, can we detect evidence of this abnormality using positron emission tomography?

One Target (Among Several…)

Rationale: Preclinical models (depletions, chronic stress) eliciting downregulation of mesolimbic DA lead to reduced DAT levels in dorsal and ventral striatum (compensatory downregulation)

(Brake et al., 2004; Lucas et al., 2004, 2007; Jiao et al. 2003).

Reduced DA synthesis/ release

DAT

1) Post-mortem: Reduced dopamine transporter in MDD in the amygdala (Klimek et al., 2002) and reduced DA turnover in the striatum (Bowden et al., 1997;). 2) In vivo molecular imaging: Inconsistent!

Dopamine Transporter (DAT) Abnormalities in MDD?

12 studies: No significant effects

• 5 studies: ↑DAT density in MDD
• 2 studies: ↓DAT density in MDD
• 5 studies: no differences

BUT:

• High heterogeneity (clinical heterogeneity?)
• 11 of 12 studies used SPECT and unspecific DAT

tracers (e.g., [123]b-CIT)

PET study with the highly selective DAT tracer 11C altropane

Dynamic PET scans acquired with an ECAT EXACT HR+ and bolus injection of [11C]Altropane

Advantages of altropane [Fischman et al 2001; Madras et al., 1998a,b,c]: 1) Rapid and specific striatal binding [max binding within 30 min in

DA-rich striatal regions, including the caudate, putamen, NAc];

2) High selectivity for DAT [28x selectivity for DAT over SERT]; 3) Low level of nonspecific binding [e.g., putamen:cerebellum: 120:1]

Pizzagalli et al., under review

Dopamine Transporter (DAT) Abnormalities in MDD?

• I. Reduced striatal DAT binding in MDD
• 0.3
• 0.2
• 0.1

0.0 0.1 0.2 0.3

Caudate Putamen NAc Binding Potential

HC MDD

*

Group x Region: Wilks’ Lambda (2,44) = 4.14 p < 0.025 [covariate: age] HC: n = 23 MDD: n = 25

Pizzagalli et al., under review

Dopamine Transporter (DAT) Abnormalities in MDD?

• II. Reduced VTA DAT binding in MDD

VTA: Main effect of Group F(1,45)= 6.04 p<0.018 [covariate: age]

• 0.04
• 0.02

0.00 0.02 0.04 0.06

Ventral Tegmental Area (VTA) Binding Potential

HC MDD

*

HC: n = 23 MDD: n = 25

Pizzagalli et al., under review

Dopamine Transporter (DAT) Abnormalities in MDD?

• III. Reduced striatal and VTA DAT binding in MDD is

exacerbated by numbers of MDE

• 0.4
• 0.2

0.0 0.2 0.4

HC 1 MDE 2-4 MDEs 5+ MDEs Binding Potential BP & # MDEs: Putamen: r = -0.36 VTA: r = -0.36 (ps<0.014, N=47)

Pizzagalli et al., under review

Dopamine Transporter (DAT) Abnormalities in MDD?

• IV. Reduced VTA DAT binding with perceived

“entrapment”

External Entrapment Scale (Gilbert et al., 1998) (// preclinical literature) “I am in a situation I feel trapped in” “I can see no way out of my current situation” “I feel trapped by my obligations” → Overlap with the concept of helplessness

r = -0.43 p<0.032

Two clinical constructs hypothesized to be linked to DA:

• Anhedonia: ns
• Feelings of Entrapment:

r = -0.43, p < 0.032

Pizzagalli et al., under review

Dopamine Transporter (DAT) Abnormalities in MDD?

• V. Reduced striatal

DAT binding in post-mortem tissue

MDD: ↓ putamen expression:

• TH (d = -1.06)
• Mature form of DAT

(80 kDa) (d = -1.15)

• Intermediate forms (50

kDa: d = -0.92; 60 kDa: d = -0.99) 15 depressed individuals (all by suicide) (38.9 y., 12 m) 15 HC (40.6 y., 13 m)

[Douglas-Bell Canada Brain Bank]

40 kDa 50 kDa 60 kDa 80 kDa Tyrosine hydroxylase

Conclusion

1) Reward learning (one form of anhedonic behavior):

• Is associated with individual differences in

frontostriatal and dopamine markers

• Can be bi-directionally perturbated by challenges

affecting striatal DA transmission (D2/3 ago/antagonists, amphetamine, nicotine, stress), including stress-induced inflammation (IL-6) [not shown, see Treadway et al., Biological Psychiatry, 2017] 2) Depression: impaired ability to modulate behavior as a function of rewards

• Dysfunctions in striatal regions implicated in reinforcement

learning (caudate) and reward prediction error (accumbens)

• DA downregulation (PET + post-mortem DAT finding)
• It can be “rescued” by a pharmacological challenge

hypothesized to transiently increase DA (amisulpride)

R01 MH095809 R01 MH101521 R01 MH102279 R01 MH108602 R37 MH068376 UH3 MH109334 Dana Foundation NARSAD Distinguished Inv. Award

Michael Treadway Roee Admon Sabina Berretta Gustavo Turecki

fMRI studies Post-mortem studies

Georges El Fakhri Rosi Kaiser

PET studies

Andre Der-Avakian Athina Markou

Preclinical studies

Acknowledgments

Dopamine Transporter (DAT) Abnormalities in MDD?

• V. Abnormal age-related DAT effects in MDD

Pizzagalli et al., under review

Pizzagalli et al., under review

Dopamine Transporter (DAT) Abnormalities in MDD?

Exaggerated Stress Responsiveness Decreased Reward Responsiveness Depression Blunted Mesolimbic DA System Biological Vulnerability Environmental Factors

Does uncontrollable stress reduce reward learning?

Three independent human studies: Healthy subjects performing the task under an acute laboratory stressor or a naturalistic stressor displayed reduced reward learning [Bogdan & Pizzagalli, 2006; Bogdan et al., 2011; Nikolova et al., 2012]

• Acute Stressor* Reduces Reward Learning

Condition: F(1,74) = 5.88 p < .020 N = 75 women

0.00 0.05 0.10 0.15 0.20 Block 1 Block 2 Block 3

Response Bias

No-stress Stress

Bogdan and Pizzagalli, Biol. Psychiatry, 2006

Condition: F(1,78) = 5.39 p < .030 N = 80 women

0.00 0.10 0.20 Block 1 Block 2 Block 3

Response Bias

Bogdan et al., J. Neurosci, 2011 * Threat-of- Shock

Ryan Bogdan

• HOW Does Stress Affect Reward Learning?
• Potential mechanisms?
• Study 1 (humans): Disruption of ventral striatal

prediction error signaling during reinforcement learning (via increased inflammatory responses)?

• Study 2 (rats): Disruption of stress peptides

(Nociceptin/orphanin FQ)?

• Study 1

Michael Treadway Roee Admon

Design:

• Healthy females (N=88)
• Stressor for behavioral session:

Maastricht Acute Stress Task (MAST; Smeets et al, 2012)

• Stressor for imaging session:

Montreal Imaging Stress Task (MIST; Dedovic et al, 2005)

• fMRI task: Instrumental

Reinforcement Task (Pessiglione et al, 2006)

• Session 1: Behavior (MAST)

Smeets et al., 2013

Time: F(2, 92) = 17.89 p = 8.0 x 10-6 N = 88 women Time (quadratic): All Fs > 66 ps < 10-12

• Session 2: fMRI (MIST)

Dedovic et al., 2005

Time (quadratic): All Fs > 24 ps < 10-6

Outcome Phase Reward Prediction Error

Pessiglione et al., Nature, 2006

fMRI Task

• Session 2: fMRI

r = -0.36 p < 0.05

Larger increases in IL-6 following stress (session 1) ↔ Larger decreases in NAcc RPE following stress (session 2)

NAc r = -0.42 p < 0.01

Left N. Accumbens RPE Change (PostStress – PreStress) IL-6 Change (PostStress – PreStress) Treadway et al., Biol. Psychiatry, 2017 Admon et al., J. Neurosci, 2017