Mass Spectrometry in Cancer Diagnostics

Postdoctoral FellowThe George L. Wright Jr. Center for Biomedical Proteomics,

Department of Microbiology and Molecular Cell Biology

Eastern Virginia Medical SchoolDiscovery Laboratory

EVMS

Gunjan Malik, Ph.D.

Eastern Virginia Medical School Norfolk, VA

NCI Early Detection Research NetworkEVMS Biomarker Discovery Laboratory

O. John Semmes, P.I.

• Prostate• Head and Neck• Breast• Bladder• Leukemia

Proteomic Profiling of CANCERS

•Ovarian•Colon •Liver•Lung

The mission of the Center is to discover a means to detect and accurately diagnose a variety of cancers long before the disease becomes life-threatening. Using mass spectrometry-based techniques to visualize and identify proteins, we hope to discover biomarkers that will detect cancer at its earliest stage, as well as markers that can predict disease and treatment outcome.

Some Facts: Why proteomics ?• More than 98% of all diseases are caused by

multiple molecular alterations.• There is very low correlation between mRNA

abundance and protein level.(e.g., yeast – S. cerevisiae: <0.5 correlation factor)

• Gene products are modified by– complex gene interactions– cellular events– environmental influences (co- & post-translational modifications: >189 types)

Pieces of the Puzzle

• Data acquisition, pre-processing, processing, signal versus noise

• Data interpretation phase 1, which signals are potential biomarkers

• Determining sources of variability, response to bias, analytical reproducibility

• Data interpretation phase 2, biomarker maturation

Ciphergen ProteinChip® Technology

SELDI ProteinChip® analysis is a mass spectrometric technology that accomplishes

protein separation on a chromatographic chip surface by binding subsets of proteins in

complex mixtures such as serum. Differentially expressed proteins are

determined by comparing protein peak intensity within mass spectra.

PHENOMICS DISCOVERY PLATFORM

The ProteinChipdiscovery platform

AssayDevelopment

Characterization& I.D.PurificationExpression

Profiling

Rapid discoveryof biomarkers:•Serum•Urine•Cell lysates•Cells•Cerebrospinal fluid•Tissue homogenates

Purification:•On-chip•Micro-chromatography spin columns•Via traditional purification strategies

Characterization andidentification through:•Epitope mapping•Peptide mapping•Phosphorylation•Peptide sequencing•Glycosylation analysis•Binding domain analysis

ProteinChip arrayis the assay.•Antibody ProteinChip for low CV’s and accurate quantity

Samples and Requirements:

Body Fluids:Serum- 20µlUrine- 60µlSeminal Plasma- 20µlCerebral Spinal Fluid- 60µl

Cell and Tissue Lysates:Microdissected cells-2000-3000Tissue Culture cells- whole cell lysatesSubcellular fractionation,

(i.e. Mitochondrial or membrane lysates)

0204060

0204060

0102030

02.55

7.5

CaP

Normal

Normal

Inte

nsity

CaP

Molecular Weight

PixCellTM

Laser Capture MicrodissectionMicroscope

Before LCM

After LCM

Captured Cells

Protein Extraction

ProteinChipTM

Surface EnhancedLaser Desorption/Ionization:Time of FlightMass SpectrometrySELDI-TOF-MS

Biomek® 2000 Laboratory Automation Workstation

Process 48 chips/384 samples per day

SELDI BioChip Arrays

100 µm x 100 µm units(10,000 per cm2)

Addressable locations

Sample spot

Focused laser beam

Laser queries multiplePositions on sample spot

20

50

80

Detector

Laser

++

+

Flight Tube

Time-Of-Flight (TOF) Mass SpectrometryLaser Desorption/Ionization

m/z

Rel

ativ

e In

tens

ityTarget:

analyte ions separated according to their mass/charge ratio

Matrix-embeddedproteins on Chip

0102030

05

1015

0102030

Gel view

SpectraHistogram

Data presentation options

An example of SELDI output20

0 pe

aks f

or se

rum

(upt

o20

0K D

a)

2500 5000 7500 10000 12500 15000

40000 60000 80000

40000 60000 8000020000

low mw

high mw

m/z

m/z

Pieces of the Puzzle

• Data acquisition, pre-processing, processing, signal versus noise

• Data interpretation phase 1, which signals are potential biomarkers

• Determining sources of variability, response to bias, analytical reproducibility

• Data interpretation phase 2, biomarker maturation

Flow diagram of spectrum analysis

SELDI-TOF MS

2-20 KDa

Baseline subtractionNormalization

Peak detection

Peak alignment and clustering

Classification (Decision Tree)

80 peaks 11000 12000 13000

10259.6 11725.2

10263.911733.5

10000

10266.511733.6

5000 10000 15000 20000

10000 11000 12000 13000

5000 10000 15000 20000

Duplicates averaged

0

2 5

5 0

7 5

0

2 5

5 0

7 5

0

2 5

5 0

7 5

0

2 5

5 0

7 5

0

2 5

5 0

7 5

0

2 5

5 0

7 5

0

2 5

5 0

7 5

0

2 5

5 0

7 5

0

2 5

5 0

7 5

0

2 5

5 0

7 5

0

2 5

5 0

7 5

0

2 5

5 0

7 5

Nor 1

Nor 2

Nor 3

Nor 4

Nor 5

Nor 6

PCA 1

PCA 2

PCA 3

PCA 4

PCA 5

PCA 6

2000 1000080004000 6000

SELDI Profiles of Normal vs. Cancer Serum Proteins

Step 1: Discovery Step 2: Evaluation Step 3: Class prediction

Training data set

Pattern discoveryx

y

Test data set

Use biomarker pattern for step 2.

x

y

Cluster analysis

Determination of:• Sensitivity• Specificity• Positive predictive value• Negative predictive value

Unknown data set

Profile 1 Profile 2

**

Disease Normal Disease Normal

Profile 1 Profile 2

**

Profile 1 Profile 2

**

x

y

Cluster analysis

Disease Normal

Pieces of the Puzzle

• Data acquisition, pre-processing, processing, signal versus noise

• Data interpretation phase 1, which signals are potential biomarkers

• Determining sources of variability, response to bias, analytical reproducibility

• Data interpretation phase 2, biomarker maturation

Validation Team

EVMS

UAB

USUHSCPDR

JHMIDMCC

UTHSCSA

NCIFCRC

UPCI

5000 7500 10000 12500

EVMS

UAB

UPitt

CPDR

JHU

CTRC San Antonio

0

10

20

30

5909.7

7759.5

9270.3

0

5

10

5909.8

7771.4 9292.8

0

20

40

5909.8

7772.09295.6

0

20

40

5905.2

7765.39287.1

0

10

20

30

5906.6

7772.09298.4

0

20

40

60 5909.17771.4

9294.5

Table 2b Inter-Lab variabilityMass Intensity S/N Resolution

Peak 1 average 5906.47 26.57 163.06 460.73stdev 6.70 9.67 107.72

CV 0.0011 0.36 0.23Peak 2 average 7768.61 35.94 242.75 505.54

stdev 8.41 6.25 82.77CV 0.0010 0.17 0.16

Peak 3 average 9289.18 30.96 244.03 439.28stdev 9.89 4.70 77.35

CV 0.0011 0.15 0.18

Semmes OJ and EPSIC members. Clin Chem. 2005 Jan;51(1):102-12.

SELDI-Prostate Cancer Validation • Phase I: Synchronization of SELDI Instruments and

Validation of Robotic Sample Processing: Can each site synchronize their instruments and procedures to match other sites?

• Phase II: Population Validation: Can multiple sites get the same result in a geographically diverse cross-section study?

• Phase III: Clinical Validation: Can multiple sites get the same result in a True Early Detection Analysis?

(STATUS: Successful Completion last year)

(STATUS: Initiating this year)

Pieces of the Puzzle

• Data acquisition, pre-processing, processing, signal versus noise

• Data interpretation phase 1, which signals are potential biomarkers

• Determining sources of variability, response to bias, analytical reproducibility

• Data interpretation phase 2, biomarker maturation

Purification

400 600 800 1000 m/z0

20

40

60

80

100

Rel

ativ

e A

bund

ance

1029.4

707.7

1030.3918.4708.8 836.6

1011.5673.7 746.5504.4 631.5405.4362.4

MS/MS

SELDI-TOF

Summary of Biomarker Discovery and Identification

Identification

Nor

mal

ized

Inte

nsity

8500 9000 9500 m/z

Classification and Regression Tree Analysis

[Cancer Research 62, 3609-3614, July 1, 2002] © 2002American Association for Cancer Research

8216777

7819.75 ? 0Yes

Yes

No

No

BPH

PCA

23270

801357

1569

78540

2817

1271

7024.02 ? 0 9149.12 ? 1.5404

0057

5382.13 ? 0

1512

7460

4480

004

2813

BPH PCA

9655.75 ? 0.1912 9507.61 ? 1.4318

Yes No

Yes No Yes NoYes No

NormalBPH

151

0011

040

7420

0460

420

PCAPCAPCA Normal

4474.71 ? 2.7869 8141.23 ? 1.3710 5074.16 ? 4.8813Yes No Yes No Yes No

L1

L2

L3 L4 L5 L6 L7 L8

L9 L10

8216777

7819.75 ? 0Yes

Yes

No

No

BPH

PCA

23270

801357

1569

78540

2817

1271

7024.02 ? 0 9149.12 ? 1.5404

0057

5382.13 ? 0

1512

7460

4480

004

2813

BPH PCA

9655.75 ? 0.1912 9507.61 ? 1.4318

Yes No

Yes No Yes NoYes No

NormalBPH

151

0011

040

7420

0460

420

PCAPCAPCA Normal

4474.71 ? 2.7869 8141.23 ? 1.3710 5074.16 ? 4.8813Yes No Yes No Yes No

L1

L2

L3 L4 L5 L6 L7 L8

L9 L10

Western SDS-PAGE

ValidationNO PCA NO PCA

1-Dimensional PAGE

• Remove Stain

• Dehydrate with 100% Acetonitrile

• Dry in Speed Vac

• Re-hydrate slice in 12.5 ng/µL Trypsin Solution

• Incubate 45 minutes on Ice

• Add Ammonium Bicarbonate to Keep Moist

• Incubate Overnight

• Extract with Formic Acid/Acetonitrile - Dry in Speed Vac

• Resuspend in Buffer “A”

Sample Preparation

2-Dimensional PAGE

For each scan in the Chromatogram, there are 3“parent” ion masses each with an associated listof fragment ions:

Peptide List (1046.7):1029.4 16410.2918.4 5211.5836.6 6187.2746.5 817.3707.7 1181.3673.7 10527.8…..

Parent Ion

Fragment Ions+

Intensity

There are ~ 2500 scans/sample with a 90 min gradient

Peptide List (1046.7):1029.4 +2918.4 +1836.6 +1746.5 +2707.7 +2673.7 +1…..

Peptide List (gi | 148566):769.2 +1740.9 +21765.3 +11634.8 +2879.5 +31028.1 +1…..

Peptide List (gi | 23997):1029.1 +2918.4 +1836.5 +1746.5 +2707.2 +2673.8 +1…..

Peptide List (gi | 428418):744.6 +11929.5 +11143.7 +21417.6 +2862.3 +31596.8 +2…..

Human NR Database

TurboSequest Search

XCorr = 0.02

XCorr = 4.73

XCorr = 0.00

400 600 800 1000 m/z0

20

40

60

80

100

Rel

ativ

e A

bund

ance

1029.4

707.71030.3918.4708.8 836.6

1011.5673.7 746.5504.4 631.5405.4362.4

IMAC-Cu2+ FPLCC-8 HPLCNO

PCa

Load

FTWash1EL-1EL-2

8000 8500 9000 9500 10000

SDS-PAGE

LC-MS/MS

BA

Apolipoprotein A II

- 14.4

- 6.0

- 3.5

PCa PCaNONO MWkDa

PCa PCa NO

6.0-

3.5-

- 10.0

14.4 - - 15.0

NOMWkDa

Malik, G. et al. Clin. Cancer Res. 2005 Feb 1;11(3):1073-85.

SELDI ProteinChip® Immunoassay

+ EAM

+ sample

wash

1 2 3 4 6 7 85

1 2 3 4 6 7 85

1 2 3 4 6 7 85

Control Ab Bait Ab

1. Capture/retention

2. Detection

3. Data analysisApoA II Titration Curve

y = 1.3198xR2 = 0.985

0

100

200

300

400

500

600

700

800

0 100 200 300 400 500 600

ApoA II Conc. (ug/ml)

M/Z

Are

a

8000 8500 9000 9500 10000

0 µg/ml10 µg/ml

25 µg/ml50 µg/ml

100 µg/ml200 µg/ml400 µg/ml600 µg/ml800 µg/ml900 µg/ml

1000 µg/ml

15 µg/ml

mass/charge

Nor

mal

ized

Inte

nsity

SELDI Immunoassay. A group of 50 NO (healthy), 51 BPH (benign) and 48 PCa (prostate cancer) samples were screened for ApoA-II levels using the SELDI-based immunoassay on mouse mAb Anti-ApoA-II coated PS20 ProteinChips®. The 8.9K m/z area observed in the serum samples was used to calculate the serum levels of ApoA-II based on the titration curve. About 4 fold increase in the protein amount in PCa samples is observed as compared to NO. (D) The relative normalized intensity observed in the SELDI-TOF-MS 8.9K m/z trace on IMAC-Cu2+ ProteinChips®.

0

20

40

60

80

100

120

140

160

Am

t. A

poA

-II (u

g/m

l)

33.5820332 151.283344 136.4786973

NO BPH PCA0

5

10

15

20

25

30

Rel

ativ

e In

tens

ity

Intensity 5.217622093 28.37109388 29.83151218

NO BPH PCa

Relative intensity on IMAC-Cu2+ Relative ApoA-II Amount

BA

C

Pure ApoA II6.0 -

kDa

PCa-loa

d

NO-Loa

d

PCa-IP

NO-IP

8700 8800 8900 9000 9100 9200

Load

Cleared

Sup

IP

IgG Elution

ApoA2 Elution

0

60

0

60

0

60

0

1.5

0

1.5

8943.5+HPCa

0

1020

30

40

5060

70

80

Ave

rage

Int

ensi

ty

Avg Intensity 10.887 1.588 60.406 11.921 74.164 8.621

NO -Load NO -IP PCa-

Load PCa-IP BPH-Load BPH-IP

ApoA II Immunodepletion

B

ApoA-II Immunohistochemistry. Panel A: Prostate cancer infiltrating under uninvolved prostate glands (thick, double-headed gray arrow) which shows minimal staining even in basal cells. The pattern of staining, which is cytoplasmic, membranous and nuclear in cancer, is accentuated slightly on the advancing edge and is variable (weak to strong). Inset shows prostate cancer demonstrating scattered nuclear (thin, single-headed arrows) staining. Panel B: Prostate cancer infiltrating above uninvolved prostate glands (thick, double-headed gray arrow) which shows minimal focal staining. The pattern of staining in the prostate cancer is variable. Panel C: Weak to strong staining demonstrated in the luminal cells of PIN and prostate cancer (thin, single-headed arrows) present above uninvolved prostate glands with little staining (thick, double-headed gray arrow). Original magnification X400.

A

C

SELDI-Immunoassay for Prediction of Biopsy Outcome

Ability of ApoA-II to discriminate controls from cases with-

Marginal Clinical Symptoms (DRE+/PSA- or DRE-/PSA+) PSA<4.0, Biopsy Performed• 46 Controls in which the biopsy was negative• 46 Cases in which the biopsy was positive

Each sample was randomized in duplicate and analyzed on antibody coated PS20 chips

0

20

40

60

80

100

120

140

160

180

Am

t. A

poA

-II u

g/m

l

CONTROL CASE

Amt. ApoA-II

Anti-ApoA-II ImmunoassaySELDI-Immunoassay for Prediction of Biopsy Outcome

p value = 0.000015ROC Area = 0.672495

Group Index M/Z STD Intensity Average Intensity Median Intensity STD # of peaks # estimated

CASE 0 5.231969 12.41333 10.6609 7.82215 92 0

CONTROL 1 6.210689 7.890178 5.8172 6.453902 92 0

In this range the ApoA-II may prove a useful biomarker for

contributing to the PSA test by complementing this marker in the range where PSA fails to

detect cancer.

WHAT NEXT?

• Facilitation of protein ID– LC-MS/MS– LIFT-MALDI TOF/TOF

• Fractionation, depletion and concentration steps– Higher resolution (MALDI TOF/TOF)– Upfront Fractionation– Depletion of abundant protein (IgY Fractionation)

• Sensitivity vs. resolution– Concentration of depleted serum– Improved instrument sensitivity

• Mass spec-assisted Immunoassay development

Bind Protein Mixture toMagnetic Beads

Wash

Unbound

Bound

Elute Bound Protein

m/z

Inte

nsity

Read in MALDITOF/TOF

Bruker ClinProt Technology ProcessClinProtRobotics

Spot elutedProteins onAnchorPlate

(396 spots)

MALDI Plate

UltraFlex

QC IMAC chip spectrum (3-10kDa)7780.061

9303.472

5915.224

3968.254 8946.2984648.592 8152.4484290.881

5347.899 6687.4753285.828 8620.239

0.0

0.2

0.4

0.6

0.8

1.0

1.2

4x10

Inte

ns. [

a.u.

]

3000 4000 5000 6000 7000 8000 9000 m/z

6000 8000 10000

0

5

10

15

4000

3285.06

3962.49

4288.634640.52

5346.18

5913.45

6687.46

7775.57

8151.848614.21

8944.52

9299.69

Bruker Instrument

Ciphergen Instrument

Rel

ativ

e In

tens

ity

m/z

5902

.8

4211

.0

9278

.1

2954

.5

4644

.4

2662

.3

5336

.0

3265

.3

4092

.6

1946

.3

1547

.4

4964

.9

7759

.4

2210

.4

6627

.7

1733

.8

6087

.8

3884

.8

3509

.3

1361

.7

8923

.8

Eluate 1\0_I8\1\1SLin

0.00

0.25

0.50

0.75

1.00

1.25

1.504x10

Inte

ns. [

a.u.

]

5904

.6

4212

.3

3266

.1

1467

.9

2663

.4

7762

.3

5337

.6

2955

.5

9282

.0

1947

.3

4093

.7

4646

.1

3885

.1

1208

.2

6629

.9

2212

.7

2557

.1

4965

.0

6432

.2

8927

.5

1706

.7

6086

.8

8138

.3

3450

.4

Eluate 1\0_O19\1\1SLin, Baseline subt.

0.0

0.5

1.0

1.5

4x10

Inte

ns. [

a.u.

]

1016

.7

4212

.5

2935

.8

3266

.2

1468

.2

2607

.0

7760

.4

3956

.9

6628

.3

1208

.4

2382

.4

6430

.7

2014

.4

9124

.8

9414

.3

5903

.7

1790

.0

8907

.0

8135

.9

4644

.5

5337

.1

C18\0_O4\1\1SLin, Baseline subt.

0.0

0.5

1.0

1.5

2.0

4x10

Inte

ns. [

a.u.

]

7758

.7

4211

.2

1262

.7

6627

.1

1548

.2

6429

.1

4055

.2

8905

.0

4469

.5

9411

.6

8134

.1

9121

.1

6876

.9

2607

.4

7916

.4

6170

.3

2107

.5

5062

.3

3446

.9

4757

.7

Eluate 1\0_C13\1\1SLin, Baseline subt.

0.00

0.25

0.50

0.75

1.00

1.25

1.50

4x10

Inte

ns. [

a.u.

]

1000 2000 3000 4000 5000 6000 7000 8000 9000 10000m/z

WCX=84 peaks

IMAC=85 peaks

C18=62 peaks

WAX=80 peaks

203 total unique peaks mass range 1000-10000

IgY unbound serum1~10KDa 166 peaks

No IgY 1~10KDa 36 peaks

2368.8

4211.03265.4

1619.2

1547.4 2743.6

5902.4 8592.75335.0 9276.0

1220.63893.4

0.0

0.5

1.0

1.5

2.0

4x10

Inte

ns. [

a.u.

]

7763.6

1534.7

9282.91984.8 3887.5 8139.73163.0

4646.92512.3

6630.31240.7 5906.7

0.0

0.5

1.0

1.5

2.0

2.5

4x10

Inte

ns. [

a.u.

]

1000 2000 3000 4000 5000 6000 7000 8000 9000m/z

MALDI profiling of serum IgY depletion

500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500

WHAT NEXT?• Facilitation of protein ID• Automated fractionation, depletion and

concentration steps• Sensitivity vs resolution• Mass spec-assisted Immunoassay development

Anti- #147-148

Anti-#107-109

NO

PCa

NOPCa

NOPCa

- 49.0

- 62.0

36.5 -31.0 -

66.3 -55.4 -

21.5 -

Anti- #101-102

Development of these bioamarkers may provide important contributions to multiplexed immuno-assays or antibody arrays. The observed overexpression of these markers, like ApoA-II, in PCa patients with PSA < 4.0 ng, suggests that analysis of such biomarkers in serum may extend the utility of current blood testing for PCa.

500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500500 1000 1500 2000 2500 3000 3500 4000 4500

Protein Expression Profiling with MS is reproducible, semi-quantitative, and can be validated.

MS-Profiling is an effective approach to biomarker discovery and has the potential to become a powerful analytical (diagnostic) tool.

Eastern Virginia Medical SchoolEastern Virginia Medical SchoolBiomarker Discovery LaboratoryBiomarker Discovery Laboratory

InvestigatorsJohn Semmes, Ph.D.John Davis, M.D.Jose Diaz, M.D., Ph.D.Rick Drake, Ph.D.Christine Laronga, M.D.Paul Schellhammer, M.D.Jeffery T. Wadsworth, M.D.

StaffDiane Brassil Lisa CazaresMaryAnn Clements Tarek Kandil Brian MainShamina MitchellMichael D. Ward

FellowsGunjan Malik, Ph.D.Alberto Corica, M.D.Daniel Holterman, Ph.D.Lining Qi, Ph.D.Cindy, Ph.D.

Biostatistics/ComputationWMRIWilliam E. Cooke, Ph.D.Dasha I. Malyarenko, Ph.D.Denis M. Manos, Ph.D.Michael W. Trosset, Ph.D.Eugene R. Tracy, Ph.D.

EVMS