New Developments in LC-IMS-MS Proteomic

Measurements and Informatic Analyses

Erin Shammel Baker Kristin E. Burnum-Johnson, Xing Zhang, Cameron P. Casey, Yehia M. Ibrahim, Matthew E. Monroe, Tao Liu, Brendan MacLean, Brian Pratt, Michael J. MacCoss and Richard D. Smith

Pacific Northwest National Laboratory

Why are we interested in using IMS-MS?

1. We would like to increase throughput of measurements and IMS-MS analyses are able

to detect high feature numbers with fast LC separations or no LC separations at all

2. IMS-TOF MS provides greater dynamic range of detection relative to trapping (e.g.

Orbitrap) instruments

3. IMS adds complementary information to MS measurements which helps lower false

discovery rates and separates isomers

4. Detection of structural changes in peptides/ proteins that can help characterize specific

disease states (structural biomarkers)

Introduction

p (Drift Gas)

Ffriction

Ion

Fel

Ion mobility concept

velocity is constant

v = K E

K = ion mobility

Drift Cell

E

E

in out

Drift Time

Pulse of 2 ions with

same m/z but different

shape

Different conformers are separated

in time with peak heights representing

the amount of each

Ion mobility concept

velocity is constant

v = K E

K = ion mobility

Drift Cell

Features:

• NanoESI ion source with 2 inlets for on-the-fly

calibration

• Off-axis hourglass ion funnel/accumulation trap

before IMS

• Rear ion funnel after IMS

• Segmented quadrupole for CID

• High dynamic range Agilent TOF or Q-TOF MS

IMS-MS concept ESI

Source Rear IF Trapping IF

MS

Q00

High Pressure Ion Funnel

Drift Cell

DRIFT CELL

Peptides

DNA Sugars

Drift Time Electric Field

Ion mobility concept LC (minutes) IMS (~60 ms) MS (~100 µs)

IMS MS

Elution Time Drift Time m/z

0 10 20 30 40 50 60

Elution Time (minutes)

Inte

nsi

ty

20 30 40 50 60

Drift Time (ms)

20 30 40 50 60

Drift Time (ms) 20 30 40 50 60

Drift Time (ms)

1100

100

m/z

1100

100

m/z

1100

100

m/z

Spiking

Level Non-Serum Peptide

60-min LC-

IMS-TOF MS

60-min LC-

TOF MS

100-min LC-

Velos-Orbitrap

100 pg/mL Melittin ND ND ND

100 pg/mL Dynorphin A Porcine Fragment 1-13 ND ND

1 ng/mL Des Pro Ala Bradykinin ND ND

1 ng/mL Leucine Enkephalin ND ND

10 ng/mL 3X FLAG Peptide ND

10 ng/mL Substance P

100 ng/mL [Ala92]-Peptide 6

100 ng/mL Methionine Enkephalin

8 peptides spiked in human serum

Sample analyzed using Velos-Orbitrap, TOF MS and IMS-TOF MS instruments

Spiking

Level Non-Serum Peptide

60-min LC-

IMS-TOF MS

60-min LC-

TOF MS

100-min LC-

Velos-Orbitrap

100 pg/mL Melittin ND ND ND

100 pg/mL Dynorphin A Porcine Fragment 1-13 1.7 (18) ND ND

1 ng/mL Des Pro Ala Bradykinin 21 (12) ND ND

1 ng/mL Leucine Enkephalin 23 (10) ND ND

10 ng/mL 3X FLAG Peptide 115 (8) 125 (20) ND

10 ng/mL Substance P 126 (7) 138 (18) 112 (19)

100 ng/mL [Ala92]-Peptide 6 868 (4) 848 (11) 841 (12)

100 ng/mL Methionine Enkephalin 1000 (3) 1000 (9) 1000 (10)

8 peptides spiked in human serum

Scaled abundance and (CV values)

Sample analyzed using Velos-Orbitrap, TOF MS and IMS-TOF MS instruments

Benefits of IMS drift time separation 1. Improved Sensitivity & Increase Feature Detection & Confidence

IMS-QTOF MS of Bradykinin (100 pM) QTOF MS of Bradykinin (100 pM)

Benefits of IMS drift time separation 1. Improved Sensitivity & Increase Feature Detection & Confidence

Only 3 features discerned without drift time dimension (*)

2. Utilizing IMS in databases reduces FDR due to extra dimension

Benefits of IMS drift time separation

Mass Error LC Error

Drift Time Error

Extra dimension adds confidence to LC-IMS-MS Features Matches to AMT Tag DB

100-min LC-Velos Orbitrap 39806 Features with FDR <5%

60-min LC-IMS-MS 75129 Features with FDR <5% If you only use LC and MS 36057 Features with FDR <5%

Human Serum Peptide Analysis

P+

P2+

P3+

P4+

L+ L+

L2+

P+

P2+

P3+

P4+

L2+

L+

Peptide : red Lipid : blue

Benefits of IMS drift time separation

Peptides and lipids are easily distinguished

3. Distinguish different classes of compounds and sequence isomers

RPPGFSPFR+

GPFRPRFPS+

RRGPFPSPF+

NW Chem used to model 3-D conformations

Benefits of IMS drift time separation 3. Distinguish different classes of compounds and sequence isomers

Simultaneous Targeted and Discovery

Proteomic Measurements with LC-IMS-MS

Increasing sample throughput and identifications

1. To obtain the highest number of proteins, tissue samples are usually fractionated

2. However, fractionation increases analysis time, introduces errors and causes

quantitation difficulties (peptides split between fractions)

3. Our goal is to understand how many proteins can be detected in the unfractionated samples with the Velos Orbitrap, Q-Exactive and IMS-MS platforms using a single 100 minute LC gradient

Patient derived xenograft (PDX) tissue sample

Mice engrafted bilaterally with WHIM16 breast tumor (Luminal A)

6-9 weeks

WHIM16 PDX samples for MS analyses

Pooled and cryopulverized

• Genomically characterized breast-cancer-derived xenografts (Li et al, Cell Rep. 2013, 4: 1116-30)

• Has been used for evaluation of the ischemia effect on proteomic and phosphoproteomics analysis (Mertins et al, Mol Cell Proteomics. 2014, 13: 1690-704)

• Part of the Comparison Reference (CompRef) sample (has one basal, WHIM2, and one luminal A sample, WHIM16) that has been used as the QC for both identification and quantification throughout the CPTAC proteomic characterization of the TCGA tumors (Zhang et al, Nature. 2014, 513: 382-7)

Platform identifications

QExact 10489

VOrbi 2498 8607

QExact 19096

VOrbi 11105

• MS/MS undersampling (especially for Velos) led us to use an AMT tag database approach

MS/MS Identifications at 1% FDR For Qexact and VOrbi

Platform identifications

AMT Tag Matching at <1% FDR (MS/MS Database 0.01% FDR)

• MS/MS undersampling (especially for Velos) led us to use an AMT tag database approach

• AMT tag database was built from 48 fractions

• Each unfractionated run was aligned to the database for identifications

QExact 17289

VOrbi 16666

IMS 22401

QExact 10489

VOrbi 2498 8607

MS/MS Identifications at 1% FDR For Qexact and VOrbi

QExact 19096

VOrbi 11105

Identifications using AMT tag approach

Unique Peptides Non-redundant Human Proteins with at least 2 peptide identification

A representative protein was chosen for redundant peptides (peptides that are shared between multiple proteins)

8,055 peptides and 319 proteins were detected only by IMS-MS

QExact 17289

VOrbi 16666

IMS 22401

QExact 2164

VOrbi 2070

IMS 2618

Platform sequence coverage difference

Of the 1855 proteins detected by all platforms, IMS-MS has the least number of proteins with only 1 and 2 peptide identification and the highest number of proteins with more than 4 peptide identifications (best sequence coverage)

Of the 1855 proteins detected by all platforms

# of Peptides/Protein IMS QE VOrbi

1 161 245 308

2 230 317 299

3 252 259 240

>4 1212 1034 1008

Non-redundant Human Proteins

Platform sensitivity difference

The 8055 peptides detected only by IMS-MS had lower intensities than the peptides found in all platforms (IMS more sensitive)

9832 peptides detected by all platforms

8055 peptides detected only by IMS-MS Unique Peptides

Simultaneous targeted and discovery IMS-MS approach

Protein Peptides Detected/Protein

Q-Exactive Orbitrap-Velos IMS-MS

Membrane-Associated Progesterone Receptor Component 2 (PR)

1 2 3

Kynureninase Isoform B

1 0 3

Epidermal Growth Factor Receptor (EGFR) Isoform A Precursor

0 0 3

Transmembrane & TPR Repeat-Containing Protein 3

0

0 2

MAP7 Domain-Containing Protein 1 0 0 1

Validated 5 low concentration proteins by spiking in heavy labeled peptides for each (2 peptides/protein except MAP7) Proteins were chosen due to their low abundance in previous studies and potential role in breast cancer

CPTAC_LabelFree_P6_4_17Sep14_Frodo_14... 9/17/2014 11:03:30 PM

RT: 0.00 - 99.00

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95

Time (min)

0

10

20

30

40

50

60

70

80

90

100

Rel

ative

Abu

ndan

ce

NL: 2.38E9

TIC F: FTMS + p NSI

Full ms

[400.00-2000.00] MS

CPTAC_LabelFree_P6_

4_17Sep14_Frodo_14-

04-13

RT: 0.00 - 99.00

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95

Time (min)

0

10

20

30

40

50

60

70

80

90

100

Rel

ative

Abu

ndan

ce

Targeted and discovery IMS-MS approach

Analysis of EGFR peptide (TIQEVAGYVLIALNTVER3+)

CPTAC_LabelFree_P6_4_17Sep14_Frodo_14... 9/17/2014 11:03:30 PM

RT: 0.00 - 99.00

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95

Time (min)

0

10

20

30

40

50

60

70

80

90

100

Rel

ative

Abu

ndan

ce

NL: 2.38E9

TIC F: FTMS + p NSI

Full ms

[400.00-2000.00] MS

CPTAC_LabelFree_P6_

4_17Sep14_Frodo_14-

04-13

RT: 0.00 - 99.00

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95

Time (min)

0

10

20

30

40

50

60

70

80

90

100

Rel

ative

Abu

ndan

ce

Target peptide from EGFR Protein 10 nM heavy added

CPTAC_LabelFree_P6_4_17Sep14_Frodo_14... 9/17/2014 11:03:30 PM

RT: 0.00 - 99.00

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95

Time (min)

0

10

20

30

40

50

60

70

80

90

100

Rel

ative

Abu

ndan

ce

NL: 2.38E9

TIC F: FTMS + p NSI

Full ms

[400.00-2000.00] MS

CPTAC_LabelFree_P6_

4_17Sep14_Frodo_14-

04-13

RT: 0.00 - 99.00

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95

Time (min)

0

10

20

30

40

50

60

70

80

90

100

Rel

ative

Abu

ndan

ce

2.5 nM heavy added Target peptide from EGFR Protein 10 nM heavy added

0 nM heavy added

Analysis of EGFR peptide (TIQEVAGYVLIALNTVER3+)

CPTAC_LabelFree_P6_4_17Sep14_Frodo_14... 9/17/2014 11:03:30 PM

RT: 0.00 - 99.00

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95

Time (min)

0

10

20

30

40

50

60

70

80

90

100

Rel

ative

Abu

ndan

ce

NL: 2.38E9

TIC F: FTMS + p NSI

Full ms

[400.00-2000.00] MS

CPTAC_LabelFree_P6_

4_17Sep14_Frodo_14-

04-13

RT: 0.00 - 99.00

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95

Time (min)

0

10

20

30

40

50

60

70

80

90

100

Rel

ative

Abu

ndan

ce

Target peptide from MAP7 Protein 10 nM heavy added

2.5 nM heavy added

Analysis of MAP7 peptide (TPETLLPFAEAEAFLKK3+)

0 nM heavy added

Biological importance

297 IMS-MS unique proteins found in Human Protein Atlas–Cancer database

319 proteins only detected by IMS-MS

Biological importance

60 of these unique IMS-MS proteins

ranked as ‘High’ in at least 25% of breast cancer stains

Of these 60 proteins: • All have functional roles in biological

process necessary for cancer progression

• > 26% have documented mechanistic roles in breast cancer progression in complementary literature

297 IMS-MS unique proteins found in Human Protein Atlas–Cancer database

319 proteins only detected by IMS-MS

The Hallmarks of Cancer

Hanahan D, Weinberg Robert A. Cell. 2011;144:646-74

Summary

LC-IMS-MS Measurements:

• Detected lower concentration proteins than the Velos

Orbitrap and Q-Exactive platforms

• Performed discovery and targeted studies simultaneously

• Identified hallmark of cancer proteins in unfractionated

samples, which are usually only observed in fractionated

studies

IMS-MS Informatic Analyses

Using Skyline

• Currently, all of our IMS-(CID)-MS all-ions fragmentation data is manually assigned which is very time consuming and limits our ability to analyze the benefit of IMS

• We need to understand the effect of the IMS separation on DIA identifications

• Skyline provides a rapid analysis tool to study MS and MS/MS quantitation with and without the IMS dimension

Why do we want to use Skyline to analyze our IMS-(CID)-MS all-ions fragmentation data?

4-12 Torr of Buffer Gas

Fragmentation takes place after the drift cell

IMS-(CID)-MS all-ions data acquisition

Parent Spectrum Fragmentation Spectrum

Fragments have the same drift time as precursors

Parent Spectrum (3 Peptides Observed in 1 sec LC acquisition) 2 hemepeptides identified

Fragmentation Spectrum

Heme Containing Peptides

(Heme)+

Heme Containing Peptides

No Heme

Zoomed into 617 region

Find multiple hemepeptides simultaneously by looking for heme+ ion at 617

Due to time consuming manual analyses, until now we have only studied specific fragments

Parent Spectrum (11 Peptides Observed in 1 sec LC acquisition) 3 Phosphopeptides Identified

Fragmentation Spectrum

Find multiple phosphopeptides by looking for PO3 or PO4 mass difference

Phospho-peptides

Due to time consuming manual analyses, until now we have only studied specific fragments

Concentration dependent IMS-(CID)-MS experiment

BSA Concentration in Yeast

100 pM 1 nM 5 nM 10 nM 50 nM 100 nM 500 nM 1 µM

All ions fragmentation IMS-(CID)-MS

All ions fragmentation (CID)-MS

• Spiked 8 different concentrations of tryptic BSA digest from 100 pM to 1

µM into 0.1 ug/uL yeast digest

• Analyzed the effect of IMS on the MS and MS/MS quantitation for the

DIA data

IMS-MS and IMS-(CID)-MS were alternated

CE of 29 V was chosen since it fragmented 2+ ions well

IMS-MS IMS-(CID)-MS (CE=29V)

Skyline provides the ability to perform fast DIA analyses

Good quantitation from 1 µM to 1 nM for all peptides when drift time filtering was used When drift time filtering wasn’t used, the lower concentration peptides quantitation wasn’t as linear

Drift time removes interference in MS quantitation

Another peak interferes when drift time filtering isn’t performed

5 nM 10 nM

Good quantitation from 1 µM to 1 nM for 7 of 10 peptides when drift time filtering and 10 for 10 from 1 µM to 10 nM When drift time filtering wasn’t used, many interferences caused bad quantitation

Drift time filtering removes interference in MS/MS quantitation

Baseline noise increases when drift time isn’t used

Interfering peaks cause quantitation problems when drift time isn’t used

Summary

Skyline:

• Quickly analyzed the IMS-(CID)-MS data

• Showed that IMS drift time allows better MS and MS/MS

quantitation by removing interfering peaks

• Enables further studies to better understand the effect of

IMS on other types of DIA data (such as SWATH)

Conclusions IMS-MS:

• Separates isomeric peptides

• Distinguished different classes of compounds

• Allows detection of low concentration components

• Is easily coupled with LC and other front end separations

• Data analysis is becoming automated with new programs such as

Skyline

Acknowledgements • Agilent Technologies

• National Institute of Environmental Health

Sciences of the NIH (R01ES022190)

• NIH: General Medical Sciences Proteomics

Center at PNNL (2 P41 GM103493-11) and

National Cancer Institute

• Leidos Biomedical Research, Inc.

• PNNL Laboratory Directed Research and

Development Program

• Environmental Molecular Sciences

Laboratory

Biological Separations & Mass Spectrometry Group