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.