METABOLIC FINGERPRINTING

1
METABOLIC FINGERPRINTING FOOTPRINTING
Ruthie Angelovici
2
The metabolom is very sensitive to perturbation.
The metabolom is found to be more sensitive to
perturbation than the transcriptome or the
proteome . This because the activities of
metabolic pathways are reflected more accurately
in concentration of pools of metabolites then in
concentration of the relevant enzymes (or indeed
the mRNAs encoding them).
3
What is a metabolic fingerprinting? The official
version

• Metabolic fingerprinting
• A strategy of classification of samples on the
  basis of their biological statues or origin,using
  high throughput methods, usually spectroscopy.
• Metabolic fingerprinting is involved in sorting
  datasets into categories so that conclusions can
  be drawn about classification of individual
  samples.

Kell et al., Nature Reviews/Microbiology. 2005
4
How it is done
X
extraction
Y
Spectroscopy Usually by direct injection
without any chromatography step
PCA Principle Components analysis
DFA Discrimination Functional analysis
5
What is it good for?

• Screening mutant collection to identify major
  alterations in biochemical pathways .
• Assay of mode of action of drugs or assessment of
  cytoxicity.
• Evaluation of unintended secondary changes in
  transgenic food crop.
• Disease diagnostic
• Identification of biomarkers for any relevant
  classification.

6
The advantages

• It is rapid and can be used for high throughput
  analysis.
• Avoiding process of signal assignments.
• Attention is focused on those parts of the
  spectrum that are most relevant to the question
  addressed.
• It is unbiased in detecting metabolites that
  happened to be present in the sample.

7
Disadvantages

• Substantial variability in metabolic composition
  of
• samples
• Quenching of the data set
• Detection of the metabolite responsible for the
• variability in the sample is often not
  possible.

8
Metabolic fingerprinting of WT and transgenic
tobacco plants by H1 NMR and multivariate
analysis techniqueChoi et al., phytochemistry
2004.
Case Study no.1
The goal To understand metabolic pathways
connected with the defense response against
TMV-infection
The mean fingerprint of WT and CSA tobacco
metabolomic response to TMV was preformed. Major
components contributing the discrimination are
revealed.
9
step 11H NMR spectra for aqueous fractions of
transgenic and WT plants.
WT leaves
1H NMR spectra for aqueous fractions of samples
(a) leaves of wild type plants, (b) leaves of CSA
plants, (c) veins of wild type plants, (d) veins
of CSA plants.
CSA leaves
WT veins
CSA veins
10
PCA principle component analysis
PCA is an unsupervised clustering method
requiring no knowledge of the data set structure
and acts to reduce dimensionality of multivariate
data whilst preserving most of variance within
it. Hence it terms as data compression method.
D1 ax1 bx2cx3dx4
11
Step 2 PCA analysis of the aqueous fractions
can separate the different samples
WNL wild type non-inoculated leaf, WIL wild
type inoculated leaf, WSL wild type systemic
leaf, CNL CSA non-inoculated leaf, CIL CSA
inoculated leaf, CSL CSA systemic leaf, WNV
wild type non-inoculated vein, WIV wild type
inoculated vein, WSV wild type systemic vein,
CNV CSA non-inoculated vein, CIV CSA inoculated
vein, CSV CSA systemic vein
12
Step 3 Identification of the discriminatory
variables
Chlorogenic acid- discriminatory variable of WNL
and WSL leaves from the rest. Glucose-
discriminate WNL and WSL veins from the rest
Alanine and malic acid -discriminate CSA plants
from the rest.
Chlorogenic acid
Malic acid
Effect of chlorogenic acid (d 7.66) and malic
acid (d 2.72) on the differentiation of aqueous
fraction of tobacco plants on the plot of PC1 and
PC2 scores. WNL wild type non-inoculated leaf,
WIL wild type inoculated leaf, WSL wild type
systemic leaf, CNL CSA non-inoculated leaf, CIL
CSA inoculated leaf, CSL CSA systemic leaf, WNV
wild type non-inoculated vein, WIV wild type
inoculated vein, WSV wild type systemic vein,
CNV CSA non-inoculated vein, CIV CSA inoculated
vein, CSV CSA systemic vein.
13
Metabolic fingerprinting of Salt stressed
tomatoesJohnson et al., phytochemistry. 2003
Case study No.2
The aim of this study was to study the effect of
salinity on tomatoes fruits Two varieties
were studied Salt tolerant tomatoes -Edkawy
(growth wise) Regular tomatoes -Sigme F1
(growth wise)
14
Representative FT-IR spectra from whole tomato
fruit flesh of Edkawy and Simge F1

• In both tomato varieties, salt treatment
  significantly reduced
• mean fruit fresh weight
• size class
• marketable yield due to BER in response to
  salinity
• no effect on total fruit number.

15
PCA analysis could not separate the two samples
Principal component analysis (PCA) models for (a)
Edkawy and (b) Simge F1 showing no discrimation
between the control fruit (0) and the fruit from
salt-treated plants (1) in either variety. (a)
Edkawy (b) Simge F1
16
DFA Discrimination Function Analysis
DFA is a supervised clustering method that
requires a priory knowledge method of replicate
structure within the data set and seeks to
minimize the within group variance and to
maximize the between group variance. The number
of principal components used by the DFA is
optimized by cross validation, which involves
forming the model on a training data set and then
projecting a previously unseen set of data , the
test set onto a model. This is a cyclical process
where the numbers of PCs are gradually reduced
to find the optimum model.
17
DFA analysis could discriminate between the two
samples
Fig 3 and 4
Fig. 3. Discriminant function analysis (DFA)
model using 20 principal components (PCs)
accounting for 99.99 total variance derived from
the raw FT-IR spectral data for Edkawy tomatoes.
Training data set contained samples 1 to 149 and
test data set contained samples 150200. The
number of PCs to be used for DFA was optimised
using the training data set and then the test
data were projected onto the DFA model. The model
shows discrimination between control and
salt-treated Edkawy tomato fruit although there
are misclassified samples in both the training
and test data sets. 0control fruit and 1salt
treated fruit.
Fig. 4. Discriminant function analysis (DFA)
model using 20 principal components (PCs)
accounting for 99.99 total variance derived from
the raw FT-IR spectral data for Simge F1
tomatoes. Training data set contained samples
190 and test data set contained samples 91120.
The number of PCs to be used for DFA was
optimised using the training data set and then
the test data were projected onto the DFA model.
The model shows discrimination between control
and salt-treated Simge F1 tomato fruit. 0control
and 1salt treated.
18
A genetic algorithm
A genetic algorithm is an optimization method
based on principles of Darwinian selection where
over a series of generation, a population of
parameters sets evolve until an optimal, or near
optimal , solution to a given problem.
19
On average total error of classification for the
GA models was below 10
fig. 6. Variable selection percentage by 50
independent genetic algorithm (GA) models, with
each model using only 5 variables, for the
discrimination between control and salt-treated
Simge F1 tomato fruit samples based on FT-IR
spectral data.
vig. 5. Variable selection percentage by 50
independent genetic algorithm (GA) models, with
each model using only 5 variables, for the
discrimination between control and salt-treated
Edkawy tomato fruit samples based on FT-IR
spectral data.
20
Conclusions
The spectral regions selected by the GA for
discrimination between control and salt treated
tomatoes are indicating a shift in biochemistry
of nitrile containing compound although further
mass spectrometric studies is required.
21
A functional genomic strategy that uses metablome
data to reveal the phenotype of silent mutations.

Raamsdonk et al., nature biotechnology 2001

• Mutants with identical phenotype should cluster
  in this plot.
• Mutant with qualitatively different phenotype
  should be clearly displaced from each other.

(1) FY23.cox5a (2) FY23.ho (3) FY23.0 (4)
FY23.pet191 (5) FY23.pfk26 (6) FY23.pfk27.
22
Metabolic footprinting

• Metabolic footprinting
• A strategy for analyzing properties of cells or
  tissues by looking in a high throughput manner at
  the metabolites that exert or fail to be taken up
  from their surrounding.
• The metabolic foot printing approach recognizes
  the significance of overflow metabolism in
  appropriate media.

23
Advantages over fingerprinting

• Rapid-Direct injection of the media
• Measuring intracellular metabolites is time
  consuming and subject to technical difficulties
  caused by rapid turnover of intracellular
  metabolites
• no need to quench metabolism and separate
  metabolites from the intercellular space.

24
High throughput classification of yeast mutants
for functional genomics using metabolic
footprinting. Allen et al., nature biotechnology
2003
Insert fig 1
25
Metabolic footprinting may be used to classify
strains on the basis of the deletion they carry.
An experiment was set up in which 24-h microtiter
plate footprints of 19 different deletant strains
with a broad range of metabolic defects were
compared. (a,b) Footprint data were used to train
a DFA model (20 PCs, 99.6 of the variance).
Footprint data from strains harboring the nit3
and pfk27 deletions clustered closely together
with strains carrying deletions in the respective
isoenzymes nit2 and pfk26. Box in a indicates
region enlarged in b. DF, discriminant function.
(c) Hierarchical cluster analysis of the data
using all 18 DFs. The scale represents the
Euclidean distance in DF space.
26
Discrimination of mode of action of antifungal
substances by use of metabolic foot
printingAllen et al., Applied and environmental
biotechnology. 2004
What this is ?
The DFA scores (1 to 3) from the analysis
illustrated in Fig. 3 were averaged according to
compound (i.e., scores for the members of each
class were averaged) and subjected to HCA. A
separation of the respiratory and nonrespiratory
inhibitors was observed in the resulting
dendrogram. Fluazinam (marked with an asterisk)
is cited as an uncoupler of oxidative
phosphorylation (17), and although it might
conceivably be regarded as a respiratory
inhibitor, it is not, of course, a respiratory
chain inhibitor, and the level of inhibition it
induces in cells growing on a fermentable carbon
source is too great to arise from the inhibition
of respiration-coupled processes alone and
therefore, this compound must inhibit other
reactions within the cell, most likely on some
proton-coupled uptake process necessary for
fermentative growth.
27
6- rules can discriminate the respiratory and non
respiratory inhibitors
GA models applied on the data set exert 6 rules
that use the combination of just three m/z ratios
to discriminate the classes. Chemical
interpretation of these peaks may lead to
identification of useful marker for understanding
the bilogycal basis of these discriminations.
28
Disease diagnostic
Hart et al., Neurological sciences. 2003
Hart et al., 2003
Developing of biomarker to identify the extent of
white matter destruction in the urine. 70
prediction for the model.
fig. 2. Multivariate analysis of 1H-NMR spectra
of urines from patients with MS (MS) or other
neurological diseases (OND) or healthy controls
(H). (A) Spectra of urines from patients with MS
(MS) or other neurological diseases (OND) or from
asymptoomatic control individuals were subjected
to discrimant analysis. The geographical
representation shows three distinct clusters with
some overlap between the MS and control clusters.
Individual samples are represented by an
asterisk. By a geometric rotation of the D2 axis
in the indicated direction, a new field to
achieve distribution of the three clusters is
over different quadrants. Using the newly defined
D2 axis, a factor plot is calculated (B)
displaying the peaks that are more prominent in
MS urines in negative direction and the
predominating peaks in ONDcontrol urines in
positive direction. (C) This representation is
based on male and female MS patients and healthy
persons, showing a significant difference between
urine profiles of males and females in MS, but
not between male and female healthy individuals.
29
Developing of biomarker to identify the extent of
white matter destruction in the urine
F- contain more N-acetyleaspartate (neural damage
marker) A-B- contain marker for demyelination,
namly choline, inositol, and inflamation like
neoterpin
Fig. 3. 1H-NMR spectroscopy of myelin-immunised
monkey urines Urines were collected twice before
and four times after EAE induction, each with
1-week interval. Of all collected urine samples,
1H-NMR spectra were recorded in triplicate and
the data were subjected to multivariate analysis.
The score plot in A shows localisation of the
pre-disease clusters in one quadrant, clearly
separated from the post-disease samples. The
variation between samples collected at one time
point is remarkably low. The factor plot
corresponding to the D2 axis of score plot A is
depicted in B. In this pseudo spectrum, the peaks
in positive direction correspond to compounds
that predominate in samples A to E peaks in
negative direction correspond to compounds
predominating in samples F. It can be seen that
the spectra from samples in cluster F contain
relatively more (than average for the whole set)
of compounds in the 0.53.5 part of the spectrum.
Peaks of known urine biomarkers of MS have been
tentatively identified, namely choline (3.19 and
3.94 ppm), inositol (3.28 and 4.10 ppm),
neopterine (4.34, 4.44, 4.60, 4.70 and 5.20 ppm),
NAA (2.05 and 2.51 ppm). Panel C depicts two
representative spectra from the same animal. Some
differences can indeed be seen in the suspected
region.Fig 3
30
Conclusions

• Metabolic footprinting is a convenient ,
  reproducible and high throughput way for genome
  -wide, physiological-level characterization of
  microorganisms.
• It can have applications on metabolic
  engineering.
• Footprinting can also contribute strongly to
  testing mathematical models of cell behavior.
• Construction of metabolic models.