Metabolomics in the study of CNS Disorders: Early Lessons Learned

1
Metabolomics in the study of CNS Disorders Early
Lessons Learned

• Rima Kaddurah-Daouk,Ph.D.
• NCI, Oct 2005
• Duke University Medical Center

2
CNS Disorders

• CNS disorders are poorly understood with no
  effective therapies
• Include
• Neurodegenerative (AD, HD, PD and MND)
• Psychiatric (depressive disorders, schizophrenia,
  ADHD, addictive, and more)
• Genetic and Environmental factors
• Several hypothesis few biochemical pathways
  mapped
• Disease symptoms disease progression and response
  to therapy varies
• Animal models are not optimal
• Reliable biomarkers would help

3
The Human CNS A Hub of Metabolic Activity
4
The Electro metabolomic Brain ?
5
Preparing for the big battle in the future
Doraiswamy PM, 2003
6
A global approach for the study of CNS disorders

• The scale up of data production and analysis in
  neuroscience presents great promise for
  understanding the complexities of the brain
  (Nature Neuroscience, Vol 7, Number 5, 2004)
• Genomics, comparative genomics, gene expression
  atlases, proteomics and imaging data are starting
  to build at a significant rate
• Adding metabolomics data is essential in this
  global effort towards understanding biological
  systems as integrated whole
• We plan to leverage these new technologies and
  system approaches towards better understanding
  and treating CNS disorders

7
Which metabolomics technology to use?

• Probably more than one
• Both Random and Targeted
• MS, NMR, EC, Lipidomics, Tracers
• It depends on disease studied and pathways
  investigated
• Hypothesis generation vs. hypothesis testing mode
• We are building programs and databases
• Will make all publicly available

8
Metabolic Pathways in Parkinson Disease
Recchia, A. et al (2004)-The FASEB Journal.
200418617-626
9
Metabolic Pathways in Drug Abuse
Cami, J. et al.(2003) N Engl J Med349975-986
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
Two examples today

• Schizophrenia and a targeted approach for
  gaining insights into mechanisms of disease and
  metabolic syndrome development with anti
  psychotics
• MND and a random metabolomics approach for
  biomarker discovery

14
Lipid profiles- signatures for schizophrenia
(SCH) and for anti psychotic drugs

• SCH is relatively common, chronic and frequently
  devastating neuropsychiatric disorder
• No diagnostic test
• Positive symptoms (delusions, hallucinations)
  negative symptoms (impaired cognition and
  emotion)
• Hypothesis include
• changes in dopamine, serotonin and glutamate
  neurotransmission
• Decreased synthesis and increased degradation of
  membrane phospholipids in certain regions of the
  brain
• Atypical antipsychotic drugs
• are thought to target serotonin 5-HT2 receptors
  and dopamine D2 receptors to lesser extent
• many associated with metabolic disorders such as
  DM and metabolic syndrome

15
Patients and drugs selected

• SCH patients selected were in acute psychotic
  episode who where medication free
• First episode patients with schizophrenia or
  patients who failed to take their prescribed
  medications
• Placed them on one of three anti psychotic drugs
• Plasma samples were collected fasting at base
  line and 2-4 weeks later
• We evaluated initially plasma samples from
• 31 first episode SCH patients off medications
• 9 patients placed on risperidone, 14 patients on
  olanzapine, 4 patients on aripiprazole
• 16 controls matched for age and gender

16
Lipid profiles and lipidomics signatures for
schizophrenia (SCH) and for anti psychotic drugs

• We report early studies to exemplify power of
  approach
• Fatty acid/lipid profile data obtained
• Lipids extracted chloroformmethanol (2/1)
  individual lipid classes separated by preparative
  thin layer chromatography lipid fractions
  scraped from plateplaced in 3N methanolic-HCl N2
  atm 100degrees C 45 min fatty acid methyl esters
  extracted in hexane 0.05butylated
  hydroxytoluene FA methyl esters were seperated
  and quantified by capillary GC
• concentration of more than 300 lipid metabolites
  within each of eight classes of lipids determined
  
• Quantitative (nmol fatty acid/g) or mole

17
Comparison of groups and visualization

• Significant differences between SCH patients and
  matched control group or upon treatment with anti
  psychotic drugs is determined
• unpaired Student's t-tests (P lt 0.05)
• Many other statistical approaches are used to
  evaluate differences
• Quantitative data is visualized using the
  Lipomics Surveyor software system
• The system creates a "heat map" graph for
  significant differences between samples
• The brightness of each individual square denotes
  the magnitude of the difference, as displayed
  with each of the heat maps
• Differences not meeting P lt 0.05 are shown in
  black
• Quantitative or mole values shown

18
Baseline vs. Control Quantitative
Nmoles/gram data, view percent difference, plt0.05
19
Baseline vs. Control Mole Percent
Mole percent data, view percent difference, plt0.05
20
Observations at Baseline

• Phosphatidylcholine and Phosphatidylethanolamine
  concentrations are down
• A clear pattern towards decrease in long chain
  polyunsaturated fatty acids suggests impairments
  in membrane structures

21
Risperidone Signature-Post vs. Pre Quantitative
Nmoles/gram data, view percent difference, plt0.05
22
Risperidone Signature Post vs. Pre Mole Percent
Mole percent data, view percent difference, plt0.05
23
Observations with Risperidone

• Risperidone significantly decreased total free
  fatty acid concentrations
• increased triglycerides
• Increased lysophosphatidylcholine and
  phosphatidylethanolamine (Lysophosphatidylcholine
  is derived from phosphatidylcholine via the
  action of phospholipases)
• Other significant changes include changes in the
  concentration of 183n3 and 183n6 across
  multiple lipid classes.

24
Olanzapine Signature-Post vs. Pre Quantitative
Nmoles/gram data, view percent difference, plt0.05
25
Olanzapine Signature Post vs. Pre Mole Percent
Mole percent data, view percent difference, plt0.05
26
Observations with Olanzapine

• Decreased total free fatty acids and Increased
  total triglycerides
• suggest that there is a problem in turnover of
  fatty acids
• Increased phosphatidylcholine and
  phosphatidylethanolamine
• Changes in the concentration of 204n3 and 203n6
  across multiple lipid classes
• From examination of the mole percent data we
  find
• composition of the free fatty acid, triglyceride,
  and phosphatidylethanolamine classes changed
  minimally
• the composition and concentration of
  phosphatidylcholine changed significantly
• production of arachadonic acid (204n6) may be
  altered by drug treatment (a desaturase enzyme
  involved ?)
• 25 lipid metabolites were significantly changed
  by both Risperidone and Olanzapine

27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
Aripiprazole Signature Post vs. Pre Quantitative
Nmoles/gram data, view percent difference, plt0.05
31
Aripiprazole Signature Post vs. Pre Mole Percent
Mole percent data, view percent difference, plt0.05
32
Observations with Aripiprazole

• Treatment with Aripiprazole had virtually no
  effect on plasma lipid metabolite concentrations
  or composition
• This drug has minimal metabolic side effects in
  schizophrenic patients

33
A Biomarker program for the study of MND

• This program brings together MGH, Metabolon,
  UPMC, NIEHS and ALSA
• Proteomics and Metabolomics technologies for
  studying perturbations in networks and pathways
  in motor neuron disease (MND)
• Focus on biomarker discovery

34
MND classification

• Heterogeneous group of rare disorders with
  diverse signs and symptoms all effecting motor
  neurons
• Terminology and classification is confusing
• Unclear as to degree of relatedness at
  biochemical and physiologic level
• Genetic susceptibility and environmental risk
  factors could contribute to disease
• Disease progression varies
• Clinical classification not precise
• ALS
• UMN
• LMN

35
Amyotrophic Lateral Sclerosis (ALS)

• The most common form of MND, large motor neurons,
  cerebral cortex, brain stem and spinal cord are
  affected
• There are familial and sporadic forms of ALS
• Results in progressive wasting and paralysis of
  voluntary muscles
• ventilatory failure and 90 death within five
  years of onset of symptoms
• No known treatment that prevents, halts or
  reverses the disease-Riluzole has marginal delay
  on mortality
• Many causes for ALS were proposed including
  glutamate excitotoxcicity oxidative stress
  mitochondrial dysfunction, autoimmune processes,
  cytoskeletal abnormalities, trophic factors
  deprivation

36
Can Metabolomics help inthe management of MND?

• Enable global understanding of changes in
  biochemical and signaling pathways in MND towards
• Better understanding of disease mechanism for new
  approaches for drug design
• Improve diagnosis of disease and its progression
• Sub classification of MND
• Enable more effective clinical trials by
  stratifying patients and providing markers for
  detecting response to therapy

37
Knowledge coming in stagesEarly days in
metabolomics

• Stage I
• Data is derived form HPLC-EC
• Powerful platform but no structures
• Simplistic in our design
• Stage II
• Preliminary data derived from GC-MS, LC-MS will
  follow
• Repeat study of samples used in HPLC-EC
• Stage III
• Collaborative program
• More sophisticated design
• Mechanistic issues

38
EC Metabolomics Platform
39
EC Chromatograms
40
Data set I

• Profile Plasma
• Simplistic design
• Selected 28 ALS and 30 controls
• Over 1500 peaks detected on HPLC-EC
• 317 metabolites were selected for analysis
• Post analysis we realized
• 23 SALS, 5LMN and 30 controls
• 16 of the MND patients were on Riluzole, 7 were
  off

41
Data mining tools

• Three measure of class association
• T-statistic
• Pearsons correlation coefficient
• Relative class association (Golub et al.)
• All three resulted in similar ranking of
  metabolites by their level of association with
  MND
• Multivariate regression used PLS-DA

42
Metabolites with significantly different
concentrations in normal and MND plasmas in the
first study
43
PLS-DA distinguished subgroups of MND in early
studies
44
Data set II

• Profile Plasma
• Still simplistic in our experimental design
• 19 ALS and 33 controls
• All MND patients were not on Riluzole
• Over 1500 peaks detected on HPLC-EC
• 317 metabolites were selected for analysis
• Post analysis we realized set included
• 13 SALS
• 4FALS (1with SOD1, 3 without)
• 2UMN
• 33 controls

45
Metabolites with significantly different
concentrations in normal controls and in MND
patients not taking Riluzole
46
PLS-DA distinguishes MND from controls in drug
fee study
47
On the identity of Two Riluzole Induced Peaks

• Selected two Riluzole induced peaks
• Major peak at 80.2 min channel 5
• Minor peak at 78.9 min channel 5
• (Riluzole peak is at 72.5 min on channel 8)
• Isolated, purified and concentrated 100 fold
• Made compatible with LC-MS
• Two peaks with retention times of 4.43 and 9.58
  min gave MS/MS fragmentation patters that were
  not related to Riluzole (Riluzole elutes at 19.9
  min)

48
Extracted Ion Chromatograms and Spectra for Peak
1 _at_4.43 min
49
Extracted Ion Chromatograms and Spectra for Peak
2 _at_9.58 min
50
a) Riluzole 2.4 ng and b) MS, MSMS, and MS3
51
Summary of Phase I Metabolomics data using
HPLC/EC platform

• Metabolic profiles based on approximately 300
  metabolites permit mathematical separation of MND
  and control subjects
• MND is associated more with a general
  down-regulation than elevation of metabolites
• A set of highly correlated metabolites is
  characteristic of a subset of MND patients that
  was significantly enriched for LMN disease
• 12 compounds are significantly up-regulated in
  MND patients taking Riluzole
• Metabolomics could help in the process of
  classification of MND
• Bigger sample sizes needed and better
  experimental design

52
Phase II MS based data in ALS metabolomics

• GC-MS and LC-MS platform
• Complex and powerful
• Key is to start the process of getting to
  structures and pathways
• Preliminary data will be shared in next set of
  slides
• GC-MS pilot study, use previous samples HPLC-EC
  to test the system
• LC-MS data in the works
• Complexities of design

53
Classification of motor neuron diseases
Classical ALS
54
Motor Neuron DiseasesHow closely are they
related as a class?
Others
PLS
FALS
ALS
UMN
HTLV ass myel
HSP
Auto immune
PMA
LMN
Kenn synd

• Factors which influence disease course are
  unknown
• age at onset, site of onset, delay from first
  symptoms to entering clinic, rate of change in
  motor and respiratory function

SMA
55
Motor Neuron DiseasesHow closely are they
related to other CNS disorders
PD
HD
Others
PLS
HTLV
UMN
HSP
AD
ALS
FALS
Auto immune
LMN
Kennedy
PMA
Others
56
Metabolic Signatures in CNSPlasma vs. CSF

• What does a signature in plasma mean?
• What is contributed from the brain?
• Does it reflect the death of a motor neuron?
• What is a result from involvement of other organs
  such as muscle?
• Are we sure effects of all drugs and
  environmental variations are sorted out of the
  equation?
• What remains as a biomarker for the disease? What
  does that mean anyway?
• How do these signatures change as a function of
  course of disease?
• What does a signature in CSF mean?
• Does CSF mimic better brain biochemistry?
• What is common between CFS and plasma signatures?

57
Samples for determining ALS Signatures and its
specificity and classification of MND
58
Conclusions

• Metabolomics has promise in helping us dissect
  MND
• This is early days
• Bigger sample sets of each MND and hence
  collaborations at a national level
• Integrating proteomic and metabolomic data in a
  system approach will be challenging and
  interesting

59
Collaborators and Consultants
Clinical
Robert H. Brown, Jr. Merit E. Cudkowicz M.
Flint Beal
INFORMATICS Steve Rozen Bruce Kristal Scientific
advisors for ALS program Jeff Rothstein Don
Cleveland Lucie Bruijn

Clinical data coordination Kristyn Newhall
Chemisty
Technology Group EC BU/VA Wayne Matson
Mikhail Bogdanov MS-Metabolon Chris
Beecher Scott Harrison Lisa Paige Corey
DeHaven Tom Barret
Paul Vouros Jimmy Flarakos
60
TWO METABOLOMIC GROUPS IN HUNTINGTONS DISEASE
61
GENOMIC METABOLOMIC CORRELLATION
GENE POSITIVE G93A MOUSE (P) vs. NORMAL PARENT
(W) FRONTAL CORTEX 71 VARIABLES
62
(No Transcript)
63
Pharmacogenomics and Metabolomics a natural fit

• Complementary data towards understanding drug
  response

64
Pharmacogenetics
The study of the role of inheritance in
individual variation in drug response phenotype.
65
(No Transcript)
66
Nature Reviews Genetics 5 669-676 (2004)
67
Identifying genes influencing polygenic drug
responses
Evans Nature, Volume 429(6990).May 27,
2004.464-468
68
Molecular diagnostics of pharmacogenomic traits
Evans Science, Volume 286(5439).October 15,
1999.487-491
69
Add metabolomics data between genotype and
phenotype
70
Identify metabolites and pathways that influence
drug response
71
PharmacogenomicsMetabolomics-The Future
The Vision The right drug, at the right dose for
every patient.
72
(No Transcript)
73
(No Transcript)
74
(No Transcript)
75
(No Transcript)
76
Metabolomics Society Organized 2004
77
Our Mission

• "The mission of The Metabolomics Society, as
  listed in its Bye-Laws, shall be to promote the
  growth and development of the field of
  metabolomics nationally and internationally to
  provide the opportunity for collaboration and
  association among the workers in that science and
  in related sciences and connections between
  academia, government and industry in the field of
  metabolomics to provide opportunities for
  presentation of research achievements and
  creation of workshops, and to promote the
  publication of meritorious research in the
  field."

78
Metabolomics Society Japan Meeting
79
Metabolomics 2006 June 24-28The Conference
Center at Harvard Medical School