a mass spectrometry platform to quantitatively profile ... · glycolysis, here at the gapdh step....
TRANSCRIPT
A Mass Spectrometry Platform to Quantitatively Prof ile Cancer Cell Metabolism from Cell Lines to Tissu esJohn M. Asara 1,2; Jason Locasale 1,2; Xuemei Yang 1; Rami Rahal 2; Norbert Perrimon 2; Lewis C. Cantley 1,2; Eric T. Wong 1 and Matthew G. Vander Heiden 3
1Beth Israel Deaconess Medical Center, Boston, MA; 2Harvard Medical School, Boston, MA; 3Massachusetts Institutes of Technology, Cambridge, MA
We demonstrate the capabilities of a metabolomics p rofiling platform that we implemented to quantitatively target 250 endogenous water soluble cellular metabolites via SRM
Mean CV= 0.12Mean R2 = 0.973
Robust and Reproducible
XIC of -MRM (163 pairs): Exp 1, 742.000/620.000 Da ID: NADP+_nega from Sample 1 (022610Qregf301) of 022610Qregf301.wiff (Turbo... Max. 2.4e5 cps.
14.5 15.0 15.5 16.0 16.5 17. 0 17.5 18.0 18.5 19. 0 19.5Time, min
0.0
1.0e4
2.0e4
3.0e4
4.0e4
5.0e4
6.0e4
7.0e4
8.0e4
9.0e4
1.0e5
1.1e5
1.2e5
1.3e5
1.4e5
1.5e5
1.6e5
1.7e5
1.8e5
1.9e5
2.0e5
2.1e5
2.2e5
2.3e5
2.4e516.88
8-10 data points
per peak
FWHH = ~9 secondsNADP+742/620
N=3, P values <0.15, t-test
~104 dynamic range
TIC: from Sample 1 (030210jl1299GFP3) of 030210jl1299GFP3.wiff (Turbo Spray) Max . 2.9e7 cps.
1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 15.0 16.0 17.0 18.0 19.0 20.0 21.0 22.0 23.0 24.0 25.0Time, min
0.0
2.0e6
4.0e6
6.0e6
8.0e6
1.0e7
1.2e7
1.4e7
1.6e7
1.8e7
2.0e7
2.2e7
2.4e7
2.6e7
2.8e7
Inte
nsity
, cps
6.26
12.754.91 10.33
3.30
16.77
2.57
14.148.2211.60
7.042.422.28
8.8022.7118.649.55
Total Ion Chromatogram
298 MRM transitions
Direct Clinical Application : Metabolomics Profiling in Cerebral Spinal Fluid (CSF) of Glioblastoma Patients at BIDMC
GBM
Advanced
disease
No GBM class I
(Neurology clinic tapped)
GBMNo GBM class II
Metabolic profile, Not tumor
size/location responsible for
clustering of GBM patients
PCA analysis of “10 Normal” versus 10 GBM patients
Cancer's insatiable appetiteLocasale, Cantley & Vander HeidenNature Biotechnology 27, 916 - 917 (2009)
It’s more than just kinase activity and genetic def ects in cancer
In order for tumor cells to grow and divide, defects in signalingpathways due to mutations, amplification, etc. need to translateto enhanced nutrient uptake and a loss of growth control.
– Cell metabolism must be altered .
Metabolic Pathway (Subway) Map
Metabolic pathways are complex but we can focus on the fewcentral and ancient pathways such as glycolysis for an initialscreen of cellular response. In addition, metabolism and cellsignaling work in synergy to grow and divide cells.
Glycolysis
Nature Immunology, 2002
Signaling and Metabolism
N=3, P values <0.15, t-test
We show a 104 dynamic range, observe mean R2 values of ~0.97across replicates and coefficient of variation (CV) values of less than0.15. Triplicate runs result in p values less than 0.15 and our platformacquires 8-10 data points per metabolite peak with a peak width of 9seconds at FWHH.
Hierarchical Clustering of Human and Drosophila Cells
Treated with 5 Anti-Metabolite Drugs
•Lots of metabolic effects
•Shows vast differences between cell types
Hierarchical clustering is carried out on the integrated peak area listsacross experimental conditions using both freeware (Cluster, dChip,
GBMNo GBM class II
(ER tapped)
As a clinical test, the cerebral spinal fluid (CSF) from 20 patients were profiledusing our metabolomics platform (10 normal and 10 with gliobastoma (GBM)).The platform was capable of clustering the normal patients from the GBMpatients using principal components analysis (PCA). For normal and GBM,two clusters were identified and only the metabolomics profile, not MRI scanscould distinguish these GBM patients as tumor size and location were similar.
Glycolysis alteration may correlate with pAKT activation
0.5
1
2
4
1 2 3
Fo
ld C
ha
ng
e (
Se
rum
Fe
d 2
0 m
in./
Sta
rve
d)
293T 8226 H929
Glycolysis and Growth Factor Signaling
'hexose-phosphate'
'glucose.1-phosphate'
'glucose.6-phosphate'
'fructose.6-phosphate'
'fructose.1,6-bisphosphate'
'dihydroxy.acetone-phosphate'
'1,3-diphopshateglycerate'
'3-phosphoglycerate'
'phosphoenolpyruvate'
'pyruvate'
'lactate'
NO AKT
signalingVery little
AKT signaling
High AKT
signaling
tAKT
actin
tAKT
actin
pAKT pAKT
tAKT
actin
pAKT
Acute Growth Factor Stimulation with Fetal Calf Serum
Increases Levels of Some Glycolytic Intermediates
N=2, R2=0.96
Analysis of the glycolysis pathway reveals that increased levels of glycolyticintermediates may correlate directly with the level of cellular pAKT levels. Forexample, H929 multiple myeloma cells rarely exhibit pAKT activity while 293Tcells signal through pAKT at high levels.
Hierarchical Clustering of CSF from 20 BIDMC Patients
•Begin to pick apart specific
metabolites contributing to
the clustering
Differences for GBM
patients #3, #10
PI3K inhibitors shrink tumors
NVP-BEZ235 – Phase II trials
Engelman et. al., Nature Medicine, 2008
FDG – PET
– Cell metabolism must be altered .
Glucose is taken up in cancer cellsat a fast rate. Tyrosine kinaseinhibitors can shut off glucoseuptake when the target is hit.
AB/Sciex
5500 Q-TRAP
New QJet ® 2 Ion Guide Q0 Q1 Qurved LINAC ® Collision Cell
New Q3 Linear Accelerator™TrapAcQuRate™ Detector
Metabolite
SelectionFragmentation Fragment
ion
Selection
Selected Reaction Monitoring (SRM)
~300 transitionsShimadzu
Prominence
UFLC
2.0mm x 15cm
275µL/min
pH=9.0, NH4+
NH2 HILIC
Platform for Targeted Endogenous Metabolite Profiling
+/-
MarkerView v1.1 PCA
analysis software Hierarchical clustering (MatLab)
KEGG pathway mapping
Courtesy of AB/Sciex
Extract metabolites with 80% methanol
From cells/CSF/tissues
MultiQuant v1.1 Peak Area
integration software
Cancer cell
Cerebral spinal fluid
One question we are asking is whether growth 1
2
Fo
lod
ch
an
ge
(T
KI/
ve
hic
le)
Tyrosine Kinase Inhibition and Glycolysis on H929 cells
hexose-phosphate
glucose.1-phosphate
glucose.6-phosphate
fructose.6-phosphate
fructose.1,6-bisphosphate
Acute Tyrosine Kinase Inhibition Results
in a Decrease of Glycolysis Pathway
Members
Kinase inhibitor drugs induce
changes in both directions on
H929 cells
MEK inhibitor
P<0.15, N=3
SUMMARY
•A quantitative metabolomics platform was implemented usin g the5500 QTRAP to profile 250 endogenous water soluble metaboli tesfrom a single 25 min. HILIC SRM experiment with pos./neg. swi tching
In order to study cancer cell metabolism, we developed a platform using theAB/Sciex 5500 QTRAP that targets 250 endogenous water soluble metabolites from300 selected reaction monitoring SRM transitions during a single LC/MS/MSexperiment and can be used with any cell or tissue source. We then integrate peakareas from Q3 TIC using both commercial and in-house developed clustering tools.
We accomplish this in a single run with a 2.0mm x 15cm amino hydrophilicinteraction chromatography (HILIC) column at pH=9.0 using positive/negative ionswitching. Since our cycle time is only 2.0 seconds, we do not use chromatographicscheduling for SRM. This results in robust and reliable data for ~200 metabolites.
across experimental conditions using both freeware (Cluster, dChip,etc.) and internally developed clustering tools programmed in MatLab.This example shows 5 anti-metabolite across a human and fly cell line.
Metabolomics Platform Can Resolve Drug Targets in Metabolic Pathways - Glycolysis
0
20
40
60
80
100
120
140
0
1
2
3
4
5
6
Rat
io IA
A/H
20
Rat
io IA
A/H
20
Glycolysis step
Iodoacetic Acid
Iodoacetic Acid
IAA
Iodoaceticacid
Data shows point of inhibition of glycolysis
– central pathway to metabolism
Our platform can resolve most metabolic members of the glycolysis pathwayin addition to other major metabolic pathways. Iodoacetic acid is a knowninhibitor of glycolysis and we can determine the point of inhibition ofglycolysis, here at the GAPDH step. The result is a buildup of glycolysisproducts up to dihydroxyacetone-phosphate and no production ofglyceraldehyde-3-phosphate or any metabolites downstream in the pathway.
Metabolite outliers from GBM patients over average of normal
pool by fold change
TCA cycle
pyruvate
IDH1/2
mutations
0
10
20
30
40
50
60
1 2 3 4 5 6 7 8 9 10
Rat
io G
BM
/No
rmal
(A
ve.)
Top Metabolite Differences in GBM CSF
Alpha Ketogluterate
Pyruvate
Cystathionine
2-oxo-4-methylthionate
CDP-nega
2-keto-isovelerate
Maleic Acid
Fumarate
Thiamine
Imidazole
Thymine
Malate
GBM
Patient #3
GBM
Patient #10
2-HG
2-HG levels
GBM#3
2-hydroxygluterate (2-HG) is converted from α-ketogluterate
when IDH1/2 is mutatedDang L, et. al. Nature. 2009, 462:739-744.
Some of the metabolites contributing to the cluster of GBM patients #3and #10 are members of the TCA (Krebs) cycle. Interestingly, GBM patient#3 shows high levels of 2-hydroxygluterate (2-HG), a metabolite producedfrom a mutation of IDH1/2 and prevalent in gliobastoma.
Hierarchical clustering of the 20 patients can help distinguish whichmetabolites contribute to the PCA clustering groups. Notice thatgroups of metabolites are different between normal and GBM patients.
Metabolomics Profiling Platform Details
•Profile and target 250 endogenous water soluble cellular metabolites
(300 SRM) covering pathways in glycolysis and metabolism-many SRM transitions based work developed by Josh Rabinowitz, Princeton Univ.
•Positive/negative ion switching within same 25 min.
LC/MS/MS run• 50 ms switching time – Fast!
• Far quicker than 4000 QTRAP
•No chromatographic scheduling for SRM• List of 300 SRM transitions (2 sec cycle time)
• 5 ms SRM dwell time/can go to 2 ms if needed
•Amino HILIC normal phase chromatography in both
Positive and Negative modes (1 column)
•(Luna NH2 2.0 mm x 15 cm, Phenomenex)
Uniqueness of our platform:
Robust and reproducible data from ~200 metabolites
40
00
QT
RA
P
55
00
QT
RA
P
Courtesy of AB/Sciex
5500 is 10X more sensitive than 4000
in SRM mode
whether growth factor signaling (primarily AKT signaling) is
directly affecting glycolysis on a short time scale
including tyrosine kinase
inhibition.
0.5
Fo
lod
ch
an
ge
(T
KI/
ve
hic
le)
BEZ235 BKM U0126
dihydroxy.acetone-phosphate
3-phosphoglycerate
phosphoenolpyruvate
lactate
PI3K only inhibitor PI3K/mTor inhibitor
0
5
10
15
20
25
30
1,3-
diph
osph
ogly
cera
te 1
3C
citr
ullin
e_13
C
hexo
se-p
hosp
hate
_13C
urid
ine_
13C
2-ox
obut
anoa
te_1
3C
myo
-inos
itol_
13C
Ger
anyl
-PP_
13C
acet
oace
tate
_13C
Imid
azol
e_13
C
3-ph
osph
o-se
rine_
13C
pant
othe
nate
_13C
gluc
ose-
6-ph
osph
ate
_13C
deox
yrib
ose-
phos
phat
e_13
C
Ure
a_13
C
tyro
sine
_13C
prol
ine_
13C
SB
P_13
C
D-e
ryth
rose
-4-p
hosp
hate
_13C
phos
phoe
nolp
yruv
ate_
13C
indo
le_1
3C
dihy
droo
rota
te_1
3C
biot
in_1
3C
D-g
luco
nate
_13C
shik
imat
e_13
C
thym
idin
e_13
C
L-ar
gini
no-s
ucci
nate
_13C
serin
e_13
C
alla
ntoa
te_1
3C
gluc
ose-
1-ph
osph
ate_
13C
xant
hine
_13C
cyto
sine
_13C
hypo
xant
hine
_13C
carn
itine
_13C
glyc
erat
e_13
C
thym
ine_
13C
5-ph
osph
orib
osyl
-1-p
yrop
hosp
hate
_13C
ribos
e-ph
osph
ate_
13C
deox
yurid
ine_
13C
6-ph
osph
o-D
-glu
cona
te_1
3C
urac
il_13
C
sn-g
lyce
rol-3
-pho
spha
te_1
3C
N-a
cety
l-glu
cosa
min
e-1-
phos
phat
e_13
C
fruc
tose
-1,6
-bis
phos
phat
e_13
C
gluc
osam
ine_
13C
AD
P-D
-glu
cose
_13C
2,3-
Dip
hosp
hogl
ycer
ic a
cid_
13C
glut
amin
e_13
C
Rat
io o
f C13
/C12
30 Minutes of C13 Glucose Flux Through 293T Cells
We quantitatively measure metabolic flux by treating cells with C13 labeled glucoseand/or glutamine and measure the destination of labeled carbon atoms via SRM.