Metabolomics at Bruker · prediction model (see Figure 3). Figure 4 highlights the workflow for...

Innovation with IntegrityMass Spectrometry, NMR

Metabolomics at BrukerMetabolomics and Lipidomics Posterbook 2018

Table of content

MetaboScapeUniting metabolomics data processing and highly confident annotation across four MS instrumental set ups: MetaboScape 4.0 1

CCSPredict: Using a Machine Learning Approach for higher Confidence in Lipid Identification 2

MRMS aXelerateUltrafast detection of drugs and metabolites in urine by Flow Injection Analysis (FIA) coupled to Magnetic Resonance Mass Spectrometry (MRMS) 3

DI-FTICR and LC-qTOF for the comprehensive profiling of myxobacterial secondary metabolomes 4

A Framework for Ultrafast Metabolic Phenotyping Utilizing Isotopic Fine Structure and Ultra-High Resolution Magnetic Resonance Mass Spectrometry 5

Analysis of removal of pesticides and drugs in waste water by plants using Flow Injection Analysis Magnetic Resonance Mass Spectrometry 6

Automated MALDI MRMS and NMR for biomarker based determination of diabetes during pregnancy 7

N-OMICSN-omics: metabolomics for N-containing metabolites 8

Metabolomcis and Lipidomics empowered by timsTOFTowards high-throughput identification and quantitation of lipids using LC-TIMS-MS/MS 9

Untargeted Metabolomics using Trapped Ion Mobility Spectrometry with Parallel Accumulation Serial Fragmentation (TIMS-PASEF) 10

Applying Trapped Ion Mobility separation (TIMS) in combination with Parallel Accumulation Serial Fragmentation (PASEF) for analysis of lipidomics samples 11

Generation of a Collisional Cross Section Library for PlantMetabolomics using Trapped Ion Mobility Spectrometry (TIMS) 12

High resolution LC-QTOF-MS based MetabolomicsUntargeted Metabolomics in Plant Biochemistry 13

Metabolomics Profiling and Identification of Cell Culture Media 14

T-ReX LC-QTOF solutionConstruction and application of a high-resolution MS/MS library including retention time information for rapid identification of endogenous metabolites 15

Metabolomics Posterbook 2018

For research use only. Not for use in diagnostic procedures.

Uniting metabolomics data processing and highly confident annotation across four MS instrumental set ups: MetaboScape 4.0

Nikolas Kessler, Wiebke Timm, Sascha Winter, Ulrike Schweiger-Hufnagel, Sven Meyer, Aiko Barsch, and Heiko NeuwegerBruker Daltonik GmbH, Fahrenheitstraße 4, 28359 Bremen, Germany

IntroductionMetabolomics approaches may be motivated in a variety of ways, pushing different criteria into foreground: speed (throughput), separation, and/or accuracy. Tailored to these prioritized criteria different instrumental set ups will fall into favor. Combining different platforms, complementing their respective strengths, ultimately closes the gap between high-throughput and in-depth analysis methods. MetaboScape4.0, including the feature extraction and ion deconvolution algorithms T-ReX 2D, 3D and 4D, integrates the processing-, dereplication- and unknown annotation-workflows for FIA-MRMS, LC-MRMS, LC-ESI-TOF, and LC-ESI-TIMS-TOF in a single software.

Conclusions•MetaboScape provides consistent

processing and annotation workflowsfor multiple instrument types withina single software solution.

•Process and explore more than a thousand measurements a single experiment.

•While using high-throughput and automated tools, deep-dives into details are supported.

MetaboScape 4.0

Fig. 1 MetaboScape 4.0 and T-ReX support different instrument types, exploiting the respective strengths of each: For the sake of separation, annotation quality, or throughput.

T-ReX 2D T-ReX 3D T-ReX 4D

Feature processing and ion deconvolution across hundreds of analysesDue to its speed, especially FIA-MRMS is suited to create experiments containing several hundreds of measurements.

The T-ReX algorithms perform featurepicking, deisotoping, and deadductingacross all these measurements. Where LC is involved, retention time alignments arecomputed to ensure solid feature tables.

The deadducting is designed to createunanimous ion interpretation across all analyses. This is even true for featuretables combined from positive andnegative polarity.

Intelligent filter parameters provide an option to account for multifactorialdesigns during feature assessment.

MetaboScape combines the key features of MS instruments in one single software

Both automated and manual tools for metabolite annotation seamlessly adapt to the annotation quality criteria each instrumental platform provides. The chromatographic dimension from LCs is a proven indicator for annotation of knowns. MRMS instruments unlock high-resolution mass accuracy and the power to resolve isotopic fine structure (IFS). Ion mobility serves as an additional indicator for annotations. TIMS-MS/MS acquisition additionally provides a high coverage of MS/MS spectra. To assess the confidence in any annotation all five criteria are reported in a concise but detailed summary, called the Annotation Quality.

Fig. 2 Boxplot of Hippuric acid intensities of 20 groups and in total 560 FIA-MRMS spectra from urine samples, measured in negative mode. Statistics and visualizations profit from experimental designs.

Fig. 3 In MetaboScape, compound identification is supported by a variety of (semi-)automated tools, which are highly integrated and enable the creation of annotations at increasing levels of specificity and confidence.

Fig. 4 Identifying the critical neutral loss of gamma-glutamyl valine with SmartFormula 3D.

In-depth analysis foreach instrument typeIn addition to the automated high-throughput annotation tools, MetaboScape presents means to dive deep into the details of MS/MS fragmentation spectra, isotopic fine structures or collisional cross sections.

Fig. 6 Assess annotation confidence based on isotopic fine structure matching.

Fig. 5 Verifying structure annotations using MetFrag1 in-silico fragmentation, the MetFrag score and intensity coverage.

Fig. 7 Comparison to predicted CCS values of lipids2 adds to annotation confidence.

4.0

Isotope Pattern

m/z

RT MS/MS

Ion Mobility

Isotopic Fine

Structure

4.0

Annotation Confidence

m/z

MS/MSTIP / IFS

CCSRT

SmartFormula

CompoundCrawler

SmartFormula

3DMetFrag Spectral

Library

AnalyteList AnalyteList

+SpectralLibrary Id

entif

ied

Com

poun

d

[1] Ruttkies et al.; J. Cheminf. 2016, 8:3[2] Zhou et al.; Anal. Chem., 2017, 89 (17)

34S

15N33S13C15N 13C33S

13C218O 13C2H

4.0

MetaboScape

1


CCSPredict: Using a Machine Learning Approach for higher Confidence in Lipid Identification

Sebastian Wehner1, Heiko Neuweger1,Ulrike Schweiger-Hufnagel1, Sven W. Meyer1, Aiko Barsch1, Nikolas Kessler1,Lucy A. Woods11Bruker Daltonik GmbH, Bremen, Germany

IntroductionTrapped ion mobility (TIMS) mass spectrometers offer new options for higher confidence in annotations of target molecules. First, with the additional separation dimension compounds co-eluting from LC columns can be separated. This can result in cleaner MS/MS spectra – crucial for any ID in lipidomics or other small molecule workflows. Moreover, TIMS enables the determination of collisional cross sections (CCS) of ions. These values are specific properties for any ion species under given conditions (type of gas, pressure, temperature). Therefore, acquired CCS values provide increased confidence in compound identification if compared to values from literature or to predicted valuesgenerated using recently developed machine learning (ML) algorithms.

Methods

timsTOF systems generate highly reproducible lipid CCS values • 10 consecutive measurements of 24 lipid ions on a timsTOF Pro showed highly reproducible CCS

values (see Figure 1). The average RSD for all lipids was 0.17%. Minimum and maximum RSD observed were 0.07% and 0.49%, respectively.

• Measurements for the same lipid standard mix on a timsTOF instrument (n=3 runs) demonstrated a high instrument to instrument reproducibility of CCS values (see Figure 2).

Lipid CCS prediction in MetaboScape for increased confidence in lipid identification• CCS values for known standards were measured on the timsTOF Pro instrument (Figure 1). These

were very similar to values predicted by a ML algorithm in MetaboScape 4.0 (Figure 3). Note that apart from LysoPC 18:1, SM 18:1, Cer 18:1, TG 18:1 the measured lipid species were not used for generating the lipid prediction model.

• CCS values match especially well for those lipid classes which were used for training the lipid prediction model (see Figure 3).

• Figure 4 highlights the workflow for high confidence lipid identification using MetaboScape 4.0 based on a real-world example.

A) Lipid extracts from two milk samples were analyzed by LC-MS/MS and LC-TIMS-MS/MS using a timsTOF. Shown are the instrument schematic and TIMS operation from ion accumulation to serial elution, where E represents electrical field and vg the gas flow.

B) Box plot of a lipid showing different abundances in two milk lipid extracts detected in ESI positive mode.

C) Assignment of molecular formula based on accurate mass and true isotopic pattern using SmartFormulaTM – C38H76NO8P.

D) MS/MS query for ESI positive mode against LipidBlast library returns PC 30:0 as likely candidate. Characteristic head group fragment (184.07 m/z) substantiated assignment as glycerophosphocholine (PC).

E) SmartFormula3DTM algorithm: Fatty acid chains C14:0 and C16:0 could be assigned based on automatic formulae assignment for characteristic fragment ions in negative mode.

F) Very low deviation (0.8 %) between measured vs. predicted CCS values (279.7 vs. 277.5) provided further orthogonal confidence for the identification of the target lipid as: PC 14:0_16:0.

Lipid Standards # 330708 Differential Ion Mobility System Suitability Lipidomix® Kit (from Avanti Polar Lipids, Inc.) were dissolved inMethanol:Water (9:1). Milk samples wereextracted using a modified Bligh & Dyer [1] method. LC separation was performed using aBruker Elute UHPLC system (20 min gradientprogram). MS data were acquired on BrukertimsTOF and timsTOF PRO instruments with andwithout TIMS separation in ESI positive andnegative autoMS/MS modes. For TIMS dataacquisition settings of 1M and 5 M Ion Charge Control were applied for timsTOF and timsTOFPro, respectively. CCS recalibration was performed using TuneMix as calibrant (322 m/z, 152.8; 622 m/z, 201.6; 922 m/z, 241.8; 1221 m/z, 279.9). The resulting data sets were processed in the novel MetaboScape 4.0 software (Bruker Daltonics) using the Time aligned Region complete eXtraction (T-ReXTM) algorithm. T-ReX3D was applied to LC-MS/MS data and T-ReX 4D to 4-dimensional LC-TIMS-MS/MS data. Statistical analysis, molecular formula annotation, MS/MS spectral library queries using LipidBlast [2, 3, 4] and prediction of CCS values were conducted in the same integrated client/server software solution. Lipid prediction is based on a support vector regression based machine learning approach described by Zhou et al. [5].

Figure 4: High confidence lipid identification workflow: A real-world example using MetaboScape 4.0

Figure 1: Highly reproducible CCS values: timsTOF ProBox plots of CCS values from 24 ion species (corresponding to 13 lipids standards from Avanti # 330708) measured in 10 consecutive timsTOFPro runs.

Conclusions• timsTOF Pro instruments

provide highly reproducible CCS values (0.17% RSD).

• CCS values can be determindedreproducibly on timsTOF and timsTOF Pro instruments (the two instruments agree on average within 1.1 Å²).

•MetaboScape 4.0 enables prediction of lipid CCS values based on machine learning.

•CCS values increase confidence in lipid annotation as orthogonal information in addition to accurate mass, true isotopic pattern, MS/MS and retention time information.

• MetaboScape provides a fully integrated solution for confident lipid annotation.

Lipidomics timsTOF Pro

A

Figure 3: CCS values measured vs. CCS values predicted in MetaboScape 4.0Average CCS values [n=10] from 24 ion species measured on timsTOF Pro are plotted vs. the CCS values predicted using a machine learning algorithm implemented in MetaboScape 4.0. The lipid class of cholesterol esters (pink) is not covered by the CCS prediction model.

Identification Confidence

MS/MSpos

CCS Prediction

Accurate m/z &

True Isotopic Pattern

A

B

MS/MSneg

Target lipid –different abundances

References[1] Bligh E. and Dyer W.; Can J Biochem Physiol 1959, 37(8):911–7

[2] Kind T. et al.; Nature Methods 2015, 10:755-758

[3] Kind T. et al.; Anal. Chem. 2014, 86(22):11024-11027

[4] Tsugawa H. et al.; J. of Cheminformatics 2017, 9(19)

[5] Zhou Z. et al.; Anal. Chem. 2017, 89(17): 9559–9566

Acknowledgements

We thank Prof. Peter Hoffmann and Dr. Mark Condina (Future Industries Institute, The University of South Australia) for providing the lipid milk extracts.

Figure 2: CCS values measured on timsTOF vs. CCS values measured on timsTOF ProCCS values from 23 ion species (corresponding to 12 lipids standards from Avanti # 330708)measured on a timsTOF [average of n=3] are plotted against the corresponding CCS values measured on timsTOF Pro [average of n=10].

C

D

E

F

CCS measured vs. predicted:Average of absolute deviation 1.86% (STDEV 1.74)

R²= 0.9683

CCS measured timsTOF Pro vs. timsTOF:Average of absolute deviation 1.10 Å² (STDEV 0.67)

R²= 0.9993

CA

14:1

Cer

(d18

:1/1

8:1)

Cho

lest

erol

est

er 1

9:0

DG

(14:

1/14

:1/0

:0)

Lyso

PC(1

8:1)

PA 1

4:1

PC 1

4:1

PE 1

4:1

PG 1

4:1

PI 1

4:1

PS 1

4:1

SM(d

18:1

/18:

1)

TG 1

8:1

220

240

260

280

300

320

340

360

380

CC

S [Å

2 ]

[M+NH4]+[M+H]+[M+Na]+

MetaboScape

2



Meropenem

Levetiracetam

Ultrafast detection of drugs and metabolites in urine by Flow Injection Analysis (FIA) coupled to Magnetic Resonance Mass Spectrometry (MRMS)

Conclusions

Metabolomics

• Drugs and their possible metabolites could be detected by FIA-MRMS very quickly.

• The presented FIA-MRMS screening workflow is much faster than a conventional workflow using UPLC-MS.

• The fast screening method could be used for quantification if internal drug standards (13C labeled compounds) are added to the sample.

Fig. 1: Flow program for FIA-MRMSmeasurements of urine samples using Elute HT

Matthias Witt1, Markus Godejohann,2 Aiko Barsch1

1Bruker Daltonik GmbH, Bremen, Germany

2Bruker BioSpin GmbH, Rheinstetten,Germany

Introduction

MethodsMass spectrometry:

Data acquisition: - scimaX MRMS with 7 T superconducting

magnet and new dynamically harmonized analyzer cell and 2 omega detection

- mass range m/z 107 – 3000- ionization: ESI(+) and ESI(-)- resolving power of 1,350,000 at m/z 200 - 28 single scans were averaged for the

final mass spectrum

Mass calibration:- external calibration with NaTFA cluster- internal recalibration with several know metabolites in urine

Sample Introduction:- FIA with a sample loop of 20uL using UPLC Elute HT. During the FIA experiment the flow was reduced to get constant signal for roughly 1.5 min (Fig 1).

Data processing:- MetaboScape 4.0

Detection of metabolites and drugs in body fluids such as plasma or urine by LCMS is a routine method in metabolomics and doping analysis. Routine UPLC-MS measurements are performed typically in 15 min. Therefore, the number of analyzed samples is highly limited. In this work, a fast method for detection of drugs and their metabolites in urine using FIA-MRMS is presented. Roughly 250 samples can be measured in 24h using this technique.

The ESI(+) and ESI(-) data after extraction of the features in MetaboScape were combined for further analysis. More than 2100 features were found for the pooled urine samples. More than 90% of these could automatically be assigned with a molecular formula. 300 drug candidates were tentatively annotated in the urine samples using an AnalyteList derived from the HMDB urine database1 with a mass error tolerance of only 0.5 ppm. The workflow is shown in Fig. 2. The detected drugs were compared with the medication of all patients. List of medication was known for all patients but without assignment to specific pooled samples. Based on medication of all patients several drugs have been found. Most of the drugs have been detected in one or a few pooled urine samples (Fig. 3a and b). Some specific drugs could be only found in positive or negative ion mode, respectively. By comparing the relative abundances of features across all samples, possible metabolites of drugs could be assigned (Fig. 4). As an example metabolites for Cefuroxime (an enteral second-generation cephalosporin antibiotic) could be detected only in the pooled urine sample 4. Several features had the same abundance pattern (here, only found in pooled sample 4) which could indicate that these compounds are metabolites of this antibiotic. The calculated molecular formula of these compounds agree with this assumption.

Samples- 6 pooled urine samples purified by SPE

using Merck LiChrolut EN SPE cartridges. Samples were extracted with MeOH and diluted 1:100 with MeOH for FIA-MRMS

- 2 Blanks (SPE purified)

- Each sample was measured in 9 replicates

Flow program for FIA-MRMS measurements

Fig. 2: Schematic (1) MRMS aXelerate workflow: a) FIA-MRMS acquisition using a scimaX MRMS, b) Data processing and evaluation using T-ReX 2D in MetaboScape 4.0, c) Generate list of molecular formula annotations including annotation qualities, d) Putative metabolite annotations using AnalyteList of known and expected compounds, e) Statistical analysis to identify features of interest, f) Optional export for advanced statistical analyses

Results

Sample:Blank1Blank2Pool1Pool2Pools3Pool4Pool5Pool6

Levetiracetam

Meropenem

Ampicillin

Lidocaine

ESI(+)ESI(+)

ESI(+) ESI(+)

HydrochlorothiazideCefurtoxim

CarboxyibuprofenAmpicillin


Ampicillin Carboxyibuprofen

Cefuroxime Hydrochlorothiazide

ESI(-)

ESI(-)

ESI(-)

ESI(-)

Fig. 3a: Bucket statistic (box plots) of tentatively annotated drugs in pooled urine samples using FIA-MRMS in positive ion mode.

Fig. 3b: Bucket statistic (box plots) of tentatively annotated drugs in pooled urine samples using FIA-MRMS in negative ion mode.

0.00min : 312.07802242m/z

0.00min : 348.04160778m/z

0.00min : 395.07855707m/z


Cefuroxime

Searching for possible metabolites of Cefuroxime

C16H17N3O7S

C15H12N2O6S

C13H16N2O5S

Fig. 4: Detection of possible metabolites of Cefuroxime in ESI(-) using Bucket Correlation Plot (left). Feature can be directly annotated with molecular formula due to very accurate mass detection.

1 Wishart DS, Feunang YD, Marcu A, Guo AC, Liang K, et al., HMDB 4.0 — The Human Metabolome Database for 2018. Nucleic Acids Res. 2018. Jan 4;46(D1):D608-17.

Ampicillin

Lidocaine

Bucket Correlation Plot:Highlighting features which show similarabundance across all samples to the one selected. Here to the one annotated as Cefuroxime (center).

MRMS aXelerate

3

www.helmholtz-hips.de

DI-FTICR and LC-qTOF for the comprehensive profiling of myxobacterial secondary metabolomes Chantal Bader1,2, Patrick Haack1,2, Fabian Panter1,2, Matthias Witt3, Daniel Krug1,2 and Rolf Müller1,21 Microbial Natural Products, Helmholtz Institute for Pharmaceutical Research Saarland (HIPS)2 German Centre for Infection Research (DZIF), Partner Site Hannover–Braunschweig3 Bruker Daltonik GmbH, Bremen

Natural products mass spectrometry

INTRODUCTIONCurrently, the investigation of myxobacterial secondary metabolism is mostly based on screening approachesusing LC-hrMS and subsequent dereplication. One of the main goals hereby is the discovery of new naturalproducts from extracts of promising microbes. In this study, we aim to evaluate the potential of DI-FTICR (FIA-MRMS) for expanding the detection range of myxobacterial metabolites. Along these lines we conducted acomparison with our LC-qTOF standard platform and characterized the chemical space covered by bothmethods. Myxococcus xanthus DK1622 is a well-described myxobacterial model organism.

References[1] Cortina N.S., Krug D. et al., Angew. Chem.

Int. Ed. Engl. 2012, 51 (3), 811–816.[2] Hoffmann M., Auerbach D. et al., ACS

Chem. Biol. 2018, 13, 273-280.[3] Bader C.D., Haack P.A. et al., manuscript

in preparation

2 6 10 14 Time [min]

LC-qTOF DI-FTICR

Full compoundspectrum

Fast Profiling

Ultra high resolution

Separate isobarics by LC

Deep metabolomemining

Retention time

Workflow

T-ReX 2D feature extraction


200 600 1000 m/z

Known compounds from DK1622 are detectableby LC-qTOF and DI-FTICR

DI-FTICR expands the chemical spaceof detectable features

Compoundfamily

Derivative LC-qTOF DI-FTICR

Myxovirescine A, B, C, G

H, L, M

Dkxanthene 574, 560, 548, 534, 544

520

Myxalamide A

B,C

Cittiline A

Plateextracts

Liquid cultureextracts

Blankextracts

3 x 3 x 3 x

A single method is not sufficient to provide a comprehensive overview of the full myxobacterial metabolome

200 400 600 800 1000 1200 14000

100

200

300

Cou

nt

DK1622 Platte DI specific features

200 400 600 800 1000 1200 14000

30

60

90

Cou

nt

200 400 600 800 1000 1200 14000

120

240

360

Cou

nt

200 400 600 800 1000 1200 14000

30

60

90

Plate DI unique features

Plate DI common features

Plate LC common features

Cou

nt

Plate LC unique features

200 400 600 800 1000 1200 14000

100

200

Cou

nt

Liquid DI specific features

Liquid DI unique features

200 400 600 800 1000 1200 14000

20

40

Cou

nt

Liquid DI common features

200 400 600 800 1000 1200 14000

20

40

60

Cou

nt

200 400 600 800 1000 1200 14000

20

40

60

Liquid LC common features

Cou

nt

Liquid LC unique features

• Common features mostly in the mz range from 300-700.• Distribution of features found in plate extracts generally shifted

to a higher mass range.

870

All DI features

Unique DI features

13

9

2561

Knownmyxobacterial

features

2798870

All DI features

7

14

3195

Unique DI features

2816

Knownmyxobacterial

features

Plate Liquid

M. xanthus DK1622 genome vs. secondary metabolome Metabolite profiles observed on platestrongly differ from liquid cultivation

Comparison of features found with DI-FTICR cultivatingin liquid medium and on plate: Both cultivation methodsshare a wide common range, but there is also a bignumber of features only produced under one of thecultivation conditions.

Unique features

liquid

Unique features

plate

3213

2780

Commonfeatures

1352


Parameters:- maXis4G- centroid mode- 2 Hz scan rate- BEH C18 column- linear gradient

Parameters:- scimaX MRMS- 28 scans in 1.5 min- 8 M- accu time 20 ms- FIA

Statistical analysis

M. xanthusDK16229.8 Mbp

?

??

Spectral networking of MS2 featuresdetected with LC-qTOF ( ) and eight mostintense features found uniquely with DI-FTICR ( ).

More than seven different compound families have already been isolated from this strain and correlated tobiosynthetic gene clusters.[1,2] Nevertheless, bioinformatic prediction shows an even bigger biosyntheticpotential of this strain on genetic level. To create an in-depth insight into the secondary metabolome of M.xanthus DK1622, we compared MS data of DK1622 after cultivation in liquid culture and on plate using ourstandard platform LC-qTOF as well as DI-FTICR.

?

?

?? ? ?

?

?

Conclusion Statistical analysis of the M. xanthus DK1622 metabolome revealed

major differences between DI-FTICR and LC-qTOF data sets. The number of features only detectable with DI-FTICR is intriguingly

high. Additionally, these features also cover a new chemical space compared

to the ones found with LC-qTOF. Cultivation of the strains in liquid and on solid medium revealed a partly

overlapping metabolome as well as large numbers of features unique toeach cultivation method.

In summary, we show that the method of cultivation as well as thechoice of analytical method have a huge impact on the detectablemetabolome.

Although standardized and simplified workflows are essential forlarge scale screening and dereplication, these results suggestimplementing complementary methods and conditions to theseworkflows in order to drastically increase the scope of such analyses.

MRMS aXelerate

4

Met

abo

lom

ics

Post

erb

oo

k 20

18


A Framework for Ultrafast Metabolic Phenotyping Utilizing Isotopic Fine Structure and Ultra-High Resolution Magnetic Resonance Mass Spectrometry

Introduction

Matthew R Lewis1, Matthias Witt 2,Nikolas Kessler2, Aiko Barsch2,Christopher J Thompson3, Jeremy K Nicholson1

1Imperial College London, London, United Kingdom; 2Bruker Daltonik GmbH, Bremen, Germany; 3Bruker Daltonics Inc., Billerica, MA

As biobanks and clinics open their doors to the promise of metabolic phenotyping for understanding population health, powerful analytical solutions capable of large-scale deployment are needed. The chemical complexity of clinical samples is traditionally measured by mass spectrometry hyphenation with GC or LC.

While powerful, the performance of these approaches is dependent on the time invested in chromatographic separation, limiting sample throughput and narrowing the metabolic coverage to only those species amenable to the chromatographic method used.

Here we present a rapid and innovative workflow utilizing Flow Injection Analysis (FIA) with ultra-high resolution magnetic resonance mass spectrometry (MRMS) achieving broad metabolic profiles in minutes. Metabolite identification is based on high mass accuracy ( 200 samples/day)

• A combination of ultra-high mass accuracy, True Isotope Pattern, and Isotopic Fine Structure ensures confident analyte ID at any level

• Provides simultaneous analysis of known and unknown metabolites

• Access compounds not readily detectable by LCMS analysis

Phenomics

Additionally, matching of extracted features to known metabolites derived from the HMDB database enabled an assignment of nearly 500 metabolites (positive and negative ESI data) –see Table 1. These included e.g. small organic acids, amino acids and nucleotides.

Remaining features which were not annotated to known metabolites were automatically assigned with a molecular formula using SmartFormula (SF) – see Table 1. Annotation qualities are provided for each results (See Figure 1: the first green bar means below 0.2 ppm mass deviation, the second green bar reports a mSigma below 50).

Figure 2 shows the excellent reproducibility of measurements. The samples A, B and C and their mixtures could be grouped perfectly in the PCA scoring plot.

Method ResultsWith an approximate acquisition time of less than 2.5 minutes, this is almost an order of magnitude faster than a traditional “fast” 20-minute LC gradient and provides a high number of detected compounds. In this study three desalted urine samples and their 1:1 mixtures were analyzed in 6 replicates and the 1:1:1 mixture of all urine samples in 30 replicates in positive and negative ion mode in less than 6 hours (132 injections). When observed, IFS directly provides the molecular formula of the unknown metabolites. FIA-MRMS workflow reveals compounds often not readily seen in LCMS workflows (see Glucose in Figure 1) that can now be observed and annotated with IFS in a single assay.

• Desalted human urine samples were analyzed on a scimaX MRMS (Bruker Daltonics) by automated FIA utilizing an Elute UHPLC in ESI +/- mode.

• Single spectral were analyzed using MetaboScape 4.0 version, Bruker Daltonics.

• Peaks belonging to the same molecular composition were de-isotoped, subsequently aligned across samples and finally adducts were combined into pseudo spectra.

• Molecular formulae were assigned to all detected ions using the SmartFormulaTM algorithm, applying metabolic profiling specific filters to elements and element ratios.

• A list of expected metabolites in human urine (HMDB Database) enabled tentative annotation of metabolites taking into account accurate mass and isotopic fine structure information (see Figure 1).

Schematic (1) MRMS aXelerate workflowa) FIA-MRMS acquisition using a scimaX MRMSb) Data processing and evaluation using T-ReX

2D in MetaboScape 4.0c) Generate list of molecular formula

annotations, including annotation qualities d) Putative metabolite annotations using

AnalyteList of known and expected compounds

e) Statistical analysis to identify features of interest

f) Optional export for advanced statistical analyses

Measurement Features Analytes with HMDB urine list Mol. Formula proposals

(SmartFormula)

ESI(+) 2953 244 2162

ESI(-) 3906 355 2961

ESI(+) and ESI(-) merged 6539 488 4834

Mass tolerance for HMDB AnalyteList based search: 0.5 ppmMass tolerance for SmartFormula search: 0.5 ppmIsotope accuracy for SmartFormula search: 200 mSigma

Figure 1:MetaboScape: T-ReX 2D extracted features from sample cohort with AnalyteList based annotation. Highlighted is a feature annotated as Glucose, a metabolite not readily detectable by LCMS.

Table 1: Summary of T-ReX 2D extracted features and corresponding annotations for analytes and molecular formulae from data acquired in ESI +, - or combined polarities.

Figure 2 Principle component analysis of multiple injections of the samples A, B and C and their mixtures in positive ion mode.

Conclusions

Pos_DevSet1_1-2-1_01_2427Pos_DevSet1_1-2-1_01_2463

Pos_DevSet1_1-2-1_01_2499

Pos_DevS

Pos_DevSet2_1-2-2

Pos_DevSet3_1-2-3_01_2431

Pos_DevSet4_1-2-4_01_2433Pos_DevSet4_1-2-4_01_2487

Pos_DevSet5_1-2-5_01_2435

Pos_DevSet6_1-2-6_01_2437

Sample A

Sample B

Sample C

1:1 Sample A/C

1:1 Sample A/B

1:1 Sample B/C

1:1:1 Sample A/B/C

MR

MS

aX

eler

ate

5


Analysis of removal of pesticides and drugs in waste water by plants using Flow InjectionAnalysis Magnetic Resonance Mass Spectrometry

Conclusions

Metabolomics

• Pesticides, drugs and their metabolites can be detected by FIA-MRMS in reduced time in plant extracts.

• Low abundant pollutants could be detected from crude extracts without purification.

• FIA-MRMS can be used to understand if pesticides and drugs can be removed from waste water.

Fig. 1: Flow program for FIA-MRMSmeasurements of plant extracts using UPLCElute HT.

Claire Villette1, Matthias Witt2,Aiko Barsch2, Dimitri Heintz1

1 University of Strasbourg, IBMP, CNRS,Strasbourg, France

2 Bruker Daltonik GmbH, Bremen, Germany

Introduction

MethodsMass spectrometry:

Data acquisition: - scimaX MRMS with 7 T superconducting

magnet and new dynamically harmonized analyzer cell and 2 omega detection

- mass range m/z 107 – 3000- ionization: ESI(+) and ESI(-)- resolving power of 1,350,000 at m/z 200

using magnitude mode- 28 single scans were averaged for the

final mass spectrum

Mass calibration:- external calibration with NaTFA cluster- lock mass calibration with compound

C12H18F12N3O6P3 (potassium adduct in positive ion mode and chlorine adduct in negative ion mode)

Sample Introduction:- FIA with a sample loop of 20 μL using UPLC

Elute HT. During the FIA experiment the flow was reduced to get constant signal for roughly 1.5 min (Fig. 1).

Data processing:- MetaboScape 4.0

Treatment of waste water is important to remove pollutants, drugs and other potentially environmentally toxic molecules before release into the nature. Constructed wetlands in France are used to treat water from small villages. Here we present the profiling of plants to understand if pesticides and drugs can be removed from waste water through an accumulation in plants or degradation by environmental factors.

Two poplars (Populus nigra) were planted - one close to the border of a pond (polluted, ZRV) and the other away from the pond (control, CTRL). Mature leaves of these two plants were collected and extracted. Each sample was analyzed in 8 technical and 3 measurement replicates. The data of the ESI(+) and ESI(-) were combined for feature analysis. More than 3,400 features have been found for the plant extract samples (Table 1). Roughly 90% of the detected features could be assigned with a molecular formula using a mass tolerance of only 0.5 ppm. More than 360 compounds have been annotated in the samples using a food data base 1 with nearly 16,000 entries using same mass tolerance of 0.5 ppm. The workflow in shown in Figure 2. Features responsible to differentiate treated and non-treated plant sample were detected by T-test and Principle Component Analysis (PCA) (Figure 3). The samples ZRV and CTRL are well separated in the PCA scoring plot and the QC sample is the center of the plot between both groups. These compounds were analyzed in detail as possible drugs and pesticides. Based on these results 4 pesticides and 4 drug candidates were found by screening versus pesticide and drug data bases (Figure 4a and 4b). The detection of these compounds was based on very accurate mass measurements and the fact that these compounds were only found in the ZRV sample and not in the control sample.

Samples- 8 technical replicates of ZRV (polluted)

samples (perfect rejection area); 8 technical replicates of CTRL (control) samples; 2 QC samples (quality control –ZRV and CTRL 1:1 mixed)

- Methanolic plant extract (stock solutions) were diluted 1:1000 in MeOH for FIA-MRMS

Flow program for FIA-MRMS measurements

Fig. 2: Schematic MRMS aXelerate workflow: a) FIA-MRMS acquisition using a scimaX MRMS, b) Data processing and evaluation using T-ReX 2Din MetaboScape 4.0, c) Generate list of molecular formula annotations including annotation qualities, d) Putative compound annotations using AnalyteList of known and expected compounds, e) Statistical analysis to identify features of interest, f) Optional export for advanced statistical analyses

Results

Fig. 4a: Bucket statistic (box plots) of detected drugs in polluted poplar samples (ZRV, green) and control poplar samples (CTRL, red) all detected in positive ion mode.

Fig. 4b: Bucket statistic (box plots) of detected pesticides in polluted poplar samples (ZRV, green) and control poplar samples (CTRL, red) all detected in positive ion mode.

Fig. 3: Statistic analysis of the plant extract samples. a) T-test results (without QC samples) and b) Principlal component analysis (red: CTRL, green: ZRV and blue: QC samples)

Measurement Features Analytes with Food DB

Mol. Formula with SF calc.

ESI(+) 2,093 100 1,876 ESI(-) 1,444 306 1,404 ESI(+) and ESI(-) merged

3,452 383 3,116

Table 1: Detected features, analytes via Food data base and molecular formula calculation of the plant extract samples using ESI(+), ESI(-) and merged data of both polarities.

0.00min : 456.16318703m/z

Procyanidin B2

0.00min : 278.13655877m/z

0.00min : 324.14205733m/z

pheophorbide a

0.00min : 631.22654389m/z

0.00min : 577.25231139m/z

Desoxyrhaponticin

Sakebiose

0.00min : 444.11307304m/z0.00min : 507.23396882m/z

Carissanol

Gravacridonediol glucoside

0.00min : 284.08968563m/z

0.00min : 902.24842299m/z

0.00min : 774.25854431m

0.00min : 482.15553559m/z

7-Methylchrysin

0.00min : 886.22999028m/z 0.00min : 322.0818

0.00min : 370.13938341m/z

0.00min : 1006.25127231m/z

0.00min : 545.22616000m/z

0.00min : 244.07115342m/z

0.00min : 410.17067716m/z

0.00min : 722.18850507m/z

0.00min : 482.30099019m/z

Mandelonitrile rutinoside 0.00min : 324.07999

0.00min : 478.12416133m/z

0.00min : 1072.69846943m/z

0.00min : 331.12669750m/z

0.00min : 568.13472338m/z

Catechin

Bucetin

0.00min : 715.21644164m/z

Capsanthone

0.00min : 302.10013966m/z

Kanokoside A

0.00min : 542.16353621m/z

a) b)

Moxisylyte

SalbutamolSalbutamol

MoxisylyteKetobemidoneKetobemidone

OxymorphoneOxymorphone

Type:CTRLZRV

Alloxydim

Benalaxyl Furalaxyl

Dimethylanilin (N.N-)Dimethylaniline

Furalaxyl

Alloxydim

Benalaxyl

Type:CTRLZRV

-100000 -80000 -60000 -40000 -20000 0 20000 40000 60000 80000 PC 1

PC1

x105

-6.0

-4.0

-2.0

0.0

2.0

4.0

6.0

PC2

Type:CTRLZRVQC

1 http://foodb.ca

MRMS aXelerate

6



Automated MALDI MRMS and NMR for biomarker based determination of diabetes during pregnancy

Franklin E. Leach III1, Christopher J Thompson2,Jeremy J Wolff2, Jacquelyn Welko1, Anushka Chelliah3, Maureen Keller-Wood3, Gary Kruppa2,Arthur S Edison1

1University of Georgia, Athens, GA; 2Bruker Daltonics Inc., Billerica, MA; 3University of Florida, Gainesville, FL

IntroductionMass spectrometry has increasingly been applied in the clinical setting due to the high and specific information content provided to researchers that enables a positive effect on patient outcomes. A different approach that eliminates the majority of sample preparation is MALDI-MS. Beyond mixing with a suitable ionization matrix, small amounts of sample (~1 uL) can be analyzed with no prior preparation or purification after clinical collection and in a high throughput fashion via MALDI automation

• Urine and serum samples were analyzed on a Bruker solariX XR at either 9.4 T or 12 T using MALDI. For urine samples, a 4 M transient with a low mass of m/z 75 was acquired.

• 12 scans were summed, leading to an average of ~40 seconds to analyze each spot.

• In selected cases (sp. serum), 2ꙍ and AMP were employed to increase RP.

ResultsMethods• For example in serum, we have been

able to identify molecular compositions that correspond to over 100 lipid species with a mass error less than 250 ppb shown in Figure 1 (left).

• Due to the ionization mechanism during MALDI, most analytes are observed as singly charged species.

• Seen in Figure 1 (right), a mass resolving power of approximately 325,000 is required to resolve the A+2 peak of a preceding phosphocholine/ phosphoethanlolamine (PC/PE) lipid species from the A or monoisotopic peak of the following PC/PE with one less double bond.

• For urine, the MALDI automation approach has resulted in the ability to directly measure the chemical complexity of over 300 clinical urine samples plus internal/external controls and blanks (480 total spots) in less than 6 hours.

• A typical spectrum is shown in Figure 2 and demonstrates the molecular complexity of this biofluid. Shown in the Figure 2 inset is a 0.10 Th wide excerpt of the spectrum illustrating the need for the increased mass resolving power.

• Detailed analysis of this large sample set was performed within Metaboscape 3.0, using the T-Rex-2D algorithm. Annotation was done with SF and matching to the urine HMDB (http://www.hmdb.ca/).

• A key feature of the analysis is the identification of patients with elevated urine glucose levels shown in Figure 3.

Conclusions

MRMS aXelerate

Sample PrepSerum and urine were collected for clinical patients and stored at -80C. After thawing at 4C, samples were mixed 1:1 with DHB matrix and 1 µL spotted onto a 384 sample AnchorChip target. No additional preparatory steps were required before mass spectrometry analysis.

Data Analysis• Data Analysis was performed in DA 5.0 and

Metaboscape 3.0 for multivariate analysis of large sample sets.

• Automated MALDI MRMS provides the opportunity to obtain complementary information that support NMR findings on large clinical sample sets with minimal sample prep.

• MetaboScape 3.0 enabled processing of MALDI-MRMS data facilitating this higher throughput profiling workflow.

• Future work will focus on further refining the approach by incorporating isotopic fine structure and MS/MS to increase confidence in the assignment of composition and structure and to correlate MS and NMR results.

Fig. 1, left) 129 molecular compositions assigned for lipid species (*). Detected as either [M+H]+ or [M+Na]+ with Mass Error < 250 ppb. Molecular composition at m/z 203.05261 detected as [M+Na]+ and assigned as C6H12O6Na (49 ppb).

Fig. 1, right) Zoom of m/z 780-790 to indicate importance of resolving power to detect and assign lipid series that overlap due to variations in saturation (mass increases of 2H) resolving power ~325,000 required to make the split indicated in the inset.

Figure. 2) MALDI MRMS analysis of 355 clinical urine samples with no sample prep beyond the addition of DHB matrix and required just under 6 hours of instrument time.

Fig. 3) The bucket at m/z 203.05261 was detected as [M+Na]+ and assigned as D-Glucose (49 ppb). Bucket statistics for this metabolite shows that high intensities of D-Glucose can only be detected in samples of diabetic patients.

1 µL was spotted on target to yield

~480 spots including samples, controls, and blanks over two

targets to produce the representative spectra shown

below.

Disease Total

Type I Diabetes 10

Type 2 Diabetes 13

Gestational Diabetes 15

Hypertension 23

Control 103

Total 164

Subjects

MRMS aXelerate

7

N-metabolites (i.e., alkaloids) are often targeted as research candidates because oftheir potential biological activities. However, discovering N-metabolites by generalmethods is extremely slow. To resolve this situation, metabolomics methods thatspecifically target N-metabolites (so-called N-omics) are required. In MS analysis,specialized metabolites are identified by their common molecular features (e.g.,ultraviolet spectrum or product ions) . However, as N-metabolites are structurallydiverse, focusing on the common molecular features is impractical in N-omics.Therefore, N-omics development requires other important features of N-metabolites.In nature, nitrogen exists as two stable isotopes: 14N (exact mass, 14.003 Da; naturalabundance, 99.63%) and 15N (15.0001 Da, 0.036%). In the MS spectrum, 15N

N-omics: metabolomics for N-containing metabolitesRyo Nakabayashi1, Kei Hashimoto1, Tetsuya Mori1, Kiminori Toyooka1, Kazuki Saito1,21 RIKEN Center for Sustainable Resource Science, 2 Chiba University

Introduction Methods

Figure 4. N-omics in the LC-FTICR-MS.Left. The labeling rate was confirmed with 15N-labeled tryptophan.Center. Identified elemental composition and its signal intensity in each organ.Right. Identified N-containing metabolite. Upper, perivine; Lower, ajmalicine.

Results

Table 1. Breakdown of liquid fertilizer.K2HPO4 15.0 mMMgSO4・7H2O 10.0 mMCaCO3 2.5 mMFeSO4・7H2O 0.7 mMEDTA・2Na 0.7 mMH3BO3 0.3 mMMnSO4・7H2O 43.7 mMZnSO4・7H2O 10.1 mMCuSO4・5H2O 10.4 mMNa2MoO4・2H2O 7.0μ MKCl 0.5 mM(15NH4)2SO4* 37.8 mMK15NO3* 98.9 mM* General reagent was used for the non labeled plants.

a b c

d e f

g h i

Plant MaterialsThe plants were individually grown in pots filled with vermiculite (Figure 2). Thepots were placed in a plant growth room under a 16/8-h light/dark cycle with anilluminance of 252–420 µmolm-2s-1 during the light period. The temperature wasmaintained at 20–25 °C. The plants were fed daily with a non- or 15N-labeledliquid fertilizer (Table 1), and watered every 2–3 days. After 8 weeks of growth,the flowers, petals, peduncles, leaves, petioles, stems, and roots of the non-and 15N-labeled Catharanthus plants were harvested and immediatelylyophilized at −55 °C. The lyophilized materials were stored at roomtemperature with silica gel. The labeling rate of 15N was approximately 95.3%.

Figure 2. Non- (left) and 15N-labeled (right) Catharanthusroseus (Equator White Eye). Theplants were purchased from ShokoScience Co., Ltd. (Japan).

Metabolite ExtractionThe freeze-dried samples were extracted in a mixer mill (MM300, Retsch) with50 μL of 80% MeOH per mg dry weight and zirconia beads. After 7 min ofmilling at 18 Hz and 4 °C, the extractions were centrifuged for 10 min and thesupernatant was filtered through an HLB μElution plate (Waters).

LC–FTICR–MS Analysis.Ultrahigh-resolution metabolome data were acquired by an FTICR–MS solariX7.0 T (Bruker Daltonics) with an ESI source. LC–FTICR–MS analysis wasperformed as previously described (Nakabayashi et al., 2013). The FTICR–MSwas controlled by the software ftmsControl 2.1.0 (Bruker Daltonics). For internalcalibration, lidcaine (250 µM in MeOH; Tokyo Chemical Industry Co. Ltd., Tokyo,Japan) was added to both solvent A (water with 0.1% formic acid, Wako PureChemical Industries Ltd.) and solvent B (acetonitrile with 0.1% formic acid,Wako Pure Chemical Industries Ltd.). The column was changed to an XselectCSH Phenyl-Hexyl (3.5 μm, 2.1 mm × 150 mm, Waters, Milford, MA, USA).

LC–FTICR–MS/MS Analysis.The MS/MS boost analysis was carried out at a collision energy of 30 V aspreviously described (Nakabayashi and Tsugawa et al., 2015).

MALDI Analysis.The extract solutions of the non- and 15N-labeled Catharanthus plants (100 µleach) were evaporated and completely dried. The extracts were re-dissolved in10 µl of 80% MeOH. Aliquots of the concentrated extract solutions (0.2 µl) weredispensed into 384-well plates and mixed with a CHCA matrix reagent solution[0.2 µl, 70 mg/ml 80% MeOH including 0.2% trifluoroacetic acid (TFA)]. Thecrystals obtained on the plate were analyzed by an FTICR–MS solariX 7.0 T(Bruker Daltonics) operated with the MALDI source.

Preparation of Frozen Section.Fresh tissues were cut about 5 mm with a razor, embedded with a compound(Surgipath FSC22: Leica Microsystems, Germany) and frozen in a -75 °Cacetone bath (Histo-Tek Pino: Sakura Finetek Japan Co.,Ltd., Tokyo, Japan).The frozen samples were placed on the cryostat specimen disk and cut with theknife blade until the desired tissue is exposed. The face of the frozen samplewas put on the adhesive film (Kawamoto’s film method) at 16 μm thickness in acryostat (CM3050S, Leica Microsystems, Germany). The section on the filmwas freeze-dried overnight at -30°C in the cryostat (Figure 3).

IMS Analysis.The freeze-dried section on the film was attached to a glass slide (ITOcoating, Bruker Daltonik GmbH) with cellophane tape. The CHCA matrixsolution (7 mg/ml 80 % MeOH including 0.2 % TFA) was sprayed onto theglass slide using ImagePrep (Bruker Daltonik GmbH) set to the defaultparameters. The freeze-dried section with the matrix was analyzed in theFTICR–MS instrument using the MALDI source. The analytical conditionswere as follows. MALDI Control: Geometry, MTP 384 ground steel; PlateOffset, 100.0 V; Deflector Plate, 200.0 V; Laser Power, 30.00%; Laser Shots,200; Frequency, 2000 Hz; Laser focus, Small; Raster Width, 30 µm. Theanalytical conditions of MALDI in the IMS analysis were identical to thosedescribed in the MALDI analysis.

MALDI-MS

CwHxNyOz

384 plate

CHCA+

extract

non-labeled 15N-labeled

CHCA+

extract

C?H?N1≤O?non-labeled

15N-labeled∆ m/z

IMS


C?H?N1≤O?non-labeled

15N-labeled∆ m/z

LC-MS

CwHxNyOz

80% MeOH


∆ m/z

15N ∆in

tens

ity

non-labeled

non-labeled15N-labeled

m/z

Xm

/z X

+15 N

15N & 13C

non-labeled CwHx14NyOzCwHx15NyOz

ultrahigh resolutionR.P. 260,000 @ m/z 400

high resolutionR.P. 66,000 @ m/z 400

isotopic ions appear as thecounterpart of 14N-containingmonoisotopic ions (N-ions).Therefore, 15N isotopic ions arepresumed as a commonfeature in N-omics. Exploitingthe fact that 15N labelingenhances the signal intensity ofnatural 15N isotopic ions, wehere establish three different N-omics analyses: 1) ultrahigh-resolution liquidchromatography–massspectrometry (LC–MS) analysis,2) high-resolution matrix-assisted laserdesorption/Ionization (MALDI)analysis, and 3) high-resolutionimaging mass spectrometry(IMS) analysis.

Figure 1. Overview of three types of N-omics.

Figure 3. Preparation of frozensection.(a) preparation of sample, (b)filling an aluminium box with thecompound, (c) centering theposition of sample, (d)embedding the sample in a bathof acetone (-75 °C), (e) setting anembedding block on to theCryostat, (f) testing to preparesection, (g) pasting the adhesivefilm onto the block, (h) cutting thesample (thickness, 16 µm) andfreezing-dry in the Cryostat, (i)freeze-dried section

C10H18N3O6SC5H10NO3C24H21NO8C4H9N2O3C20H23N2O3C9H12NO2C10H14N5O4C4H8NO4C10H14N5O5C11H16N5O3SC31H24NO5C6H14NO2C6H12N3O3C6H13N2O3C23H32NO11C20H41N2O13C32H29N2OC5H10NO4C5H12NO2SC9H13N2O6C20H23N2O3C21H25N2O2C21H25N2O2C22H29N2O4C31H27N2C17H17N4OC21H25N2O2C5H11N2O3C21H21N2O3C25H31N2O7C27H35N2O8C22H27N2O4C20H23N2O3C20H25N2OC22H29N2O3C12H16NO7C21H29N2O3C21H23N2O3C27H31N2O8C26H33N2O9C27H35N2O9C21H25N2O3C21H25N2O4C18H21N2C21H27N2O3C21H21N2O4C28H35N2O10C18H23N2OC18H23N2OC21H23N2O4C22H23N2O4C10H13N2C23H25N2O5C23H25N2O5C17H27N2O11C23H25N2O4C24H27N2O5C26H31N2O8C21H27N2O4C21H25N2O3C21H20N2O3C21H25N2O3C10H12NO4C17H21N2O4C10H8NO4

0 5 × 108

leaf

root

peta

l

pedu

ncle

petio

le

stem

resolving power 600,000

0.2

0.4

0.6

0.8

x107

non-

labe

led

le

af

1

2

3

4

5

6x108

non-

labe

led

tript

opha

n15

N-la

bele

d tri

ptop

han

2

4

6x107

5.6 6.0 6.4 6.8 7.2 7.6Time [min]

2

4

6

x108

15N

-labe

led

leaf

205.09701

206.100412

4

6

x106

205.09686

206.100481

2

3

4

x108

207.09081

208.094541

2

3

4

5

6x108

207.09100

208.094451

2

3

4

5x107

204 205 206 207 208 209 m/z

N2

MSEIC

Inte

nsity

(cnt

s)

206.09424

206.10048

1

2

3

4

5

6x107

206.093 206.096 206.099 206.102

m/z

M + 1

15N isotopic ion

Inte

nsity

(cnt

s)

NH

COOH

NH2

[M + H]+ m/z 205.09770, C11H13N2O2

NH

N

OH

MeO

O

HH

0.0

0.5

1.0

100 150 200 250 300 350

m/z

0.0

0.5

1.0

x109

400

1.5

Inte

nsity

(cn

ts)

0

1

2

3

100 150 200 250 300 350

m/z

0.0

0.5

1.0

1.5

x108

Inte

nsity

(cn

ts)

NH

HNO

OO

observed MS/MS spectrum

perivine

observed MS/MS spectrum

ajmalicine

leaf petal

root

26 (22)15 (10)

5 (5)

7 (2)

9 (9)

2 (2) 0

64 (51)

228.

1019

3

305.

1649

2

337.19079

439.

2228

7

511.

2083

7

229.

0989

1

307.

1589

0

441.

2164

9

513.

2009

7

0

2

4

6

8

0

2

4

6

8

x109

100 150 200 250 300 350 400 450 500 550m/z

600

339.18494

non-labeled leaf

15N-labeled leaf

Inte

nsity

(cnt

s)

identified as N-containing molecular formula

LC-MS (58) MALDI –MS (60)

45 13 47

C21H25N2O2 m/z 337.191

a

C13H13N6O8 m/z 381.079

b

C21H25N2O2 m/z 337.191

c

C21H19N2O3 m/z 347.139

d

C11H21N6O7 m/z 349.147

e

C13H13N6O8 m/z 381.079

f

C20H25N2O m/z 309.196

g

C21H23N2O2 m/z 335.175

h

C21H19N2O3 m/z 347.139

i

C13H13N6O8 m/z 381.079

j

C24H29N2O4 m/z 409.212

k

C24H29N2O5 m/z 425.207

l

C24H31N2O5 m/z 427.223

m

C25H33N2O6 m/z 457.233

n

C27H31N2O8 m/z 511.208

o

Figure 4. N-omics in the MALDI-MS.Left. Chromatogram of MALDI-MS in the non-labeled leaf (upper) and 15N-labeled leaf(lower).Right. Upper, Comparative analysis of identified elemental composition in three organs.Number in parentheses indicates number of uncharacterized N-containing metabolites usingtwo databases KNApSAcK and Dictionary of Natural Products, Lower, Comparative analysisof identified elemental composition in the LC-MS and MALDI-MS.

stem

root

bud

leaf

non-labeled 15N-labeledC21H2114N2O3 C21H2115N2O3 C21H2114N2O3 C21H2115N2O3

Figure 5. N-omics in the IMS.Visualization of an N-metabolite, showing the non-labeled(left) and 15N-labeled (right) data in the IMS analysis. Leftand right diagrams of each result visualize C21H2114N2O3(m/z 349.155 [M + H]+) and C21H2115N2O3 (m/z 351.149 [M +H]+), respectively. White bar indicates 1 mm (bud, leaf, andstem) and 500 µm (root).

Figure 6. Localization of N-metabolites.(a) Representative N-metabolites accumulated at different sites in the bud (a and b), leaf (c–f) and stem(g–j), and at the same site in the leaf (k–o). These metabolites were visualized at m/z value [M + H]+.White bar indicates 1 mm.

ConclusionWe have established three different analytical methods for N-omics.

KeywordsN-omics, metabolomics, FTICR-MS, LC-MS, MALDI-MS, IMS

AcknowledgementsThis work was partially supported by the Strategic International Collaborative Research Program (SICORP), JST and Japan Advanced Plant Science Network.

N-O

MIC

S

8

Met

abo

lom

ics

Post

erb

oo

k 20

18

Cesar E. Ramirez1,2, Kendra J. Adams1, Anthony Castellanos1,Alyssa Garabedian and Francisco Fernandez-Lima1,2

1Department of Chemistry and Biochemistry, Florida International University, Miami, USA2Biomolecular Science Institute, Florida International University, Miami, USA

Analytical workflowOverview

Introduction

Acknowledgments

Experimental Results

This research was supported by the National Institute of HealthAI135469 to FFL and AI045545 to FGN. Additional support wasprovided by FIU through the Advanced Mass SpectrometryFacility at the Biomolecular Sciences Institute (BSI).

Traditional lipidomic workflows based onLC-MS/MS for separation and identification(via exact masses and fragmentationpatterns) often suffer from low throughput,as long sample prep and/or separationtimes (usually with complex 2D-LCtechnologies) are required to avoidinterferences. Additionally, auto-ms/msacquisitions usually lack sensitivity becauseof low ion transmission efficiency. Recentadvances in high-res Trapped Ion MobilitySpectrometry (TIMS) such as ParallelAcquisition Serial Acquisition (PASEF)1enable identification based on retentiontimes, accurate mass and accuratecollisional cross-section (CCS) whileallowing untargeted quantitation out of asingle chromatographic run. This workdescribes the combination of thattechnology with fast sample-prep and LCsteps for the identification and quantitationof lipids out of biological matrices.

A high throughput workflow forsimultaneous lipid identificationand quantitation based on fastsample prep combined with LC,trapped ion mobility and highefficiency MS/MS

Experimental Results

ConclusionsLC-TIMS-MS/MS using PASEF® technologyallows for simultaneous, untargetedidentification and quantitation of lipids inbiological matrices, without excessive samplepreparation steps and with a simplechromatographic run.

Figure 1. Overview of the Bruker timsTOF® instrument with detail on TIMS cell.2

References1. Meier, F. et al. “Parallel Accumulation−Serial Fragmentation (PASEF):

Multiplying Sequencing Speed and Sensitivity by Synchronized Scans in aTrapped Ion Mobility Device. J. Proteome Res. 2015, 14, 5378−5387.

2. Park, M. “timsTOF MS for Life Science Problem Solving”. Genetic Engineeringand Biotechnology news. Vol. 36, No. 19. Nov. 1st, 2016.

Experimental resultsFigure 3: Ion mobility profiles for sn-isomeric phosphatidylcholine lipidsusing high resolution TIMS-MS. Notethe shift in CCS between [M+H]+ and[M+Na]+ adducts.

Table 1: sn-positional isomeridentification of K0, CCS andinterday %RSD (n=5)

1 MS2100% targets

fragmented

418 MS245% targetsfragmented

992 MS2100% targetsfragmented

0

20

40

60

80

100

0 5 10 15 20 25Run time (min)

% B

Mechanical disruption →Sonication →centrifugation

10 uM BHT (antioxidant)

+ SPLASH® mix +300 uL BuOH:MeOH

Bio material

• Dionex Acclaim 120 C1850x3.0 mm, 3 µm, 55º CFR= 0.6 mL/min

• A: 4:6 ACN/H2O, 10 mM AmForm, 0.1% FAB: 9:1 IPA:ACN, 10 mM AmForm, 0.1% FA

LC-M

S/M

SLC

-TIM

S-M

S/M

S(P

ASEF

)LC

-TIM

S-C

ID-M

S

LipidK0 (cm2V-1s-1) CCS (Å2) % RSD K0 (cm2V-1s-1) CCS (Å2)

[M+H]+ [M+Na]+

PC 18:0-16:0 0.697 ± 0.001 292.9 ± 0.5 0.23 0.683 ± 0.001 297.8 ± 0.4

PC 16:0-18:0 0.699 ± 0.002 292.3 ± 0.9 0.13 0.684 ± 0.001 297.7 ± 1.4

PC 16:0-14:0 0.731 ± 0.002 279.7 ± 0.9 0.09 0.716 ± 0.002 285.3 ± 0.9

PC 14:0-16:0 0.732 ± 0.002 279.5 ± 0.7 0.08 0.716 ± 0.003 285.3 ± 1.0

PC 18:0-14:0 0.714 ± 0.002 286.0 ± 0.9 0.12 0.701 ± 0.003 291.2 ± 1.1

PC 14:0-18:0 0.715 ± 0.003 285.7 ± 1.0 0.13 0.702 ± 0.003 291.1 ± 1.3

Figure 2: LEFT: Biological material (mosquito ovaries, bacterial cell pellets orhuman plasma) was fortified with deuterated lipids (Avanti Splash® mix) andextracted with 1-butanol/methanol 1:1, sequentially using mechanical disruptionand sonication to disrupt cell walls. Centrifugation was the only cleanup stepbefore LC injection. RIGHT: The Bruker timsTOF® instrument performance wastested under autoMS/MS, PASEF and non-selective CID acquisition modes.

Figure 4: Comparison of limits of detection (LODs) between PASEF and LC-MSmode (no TIMS) acquisition modes. Variation of mass accuracy, retention times(RT) and collisional cross section (CCS) in biological samples spiked at two levelswith 20-fold difference (relative to calibration standards).

280 290 300

Sr 0.30 V/ms

R~130R~150

R~135R~140

PC (14:0/16:0) [M+H]//[M+Na]PC (16:0/14:0) [M+H]//[M+Na]

r=1.6r=1.4

r=1.5r=1.1

R~145

r=1.0r=1.5

R~150

R~150R~135

PC (16:0/18:0) [M+H]//[M+Na]PC (18:0/16:0) [M+H]//[M+Na]

PC (14:0/18:0) [M+H]//[M+Na]PC (18:0/14:0) [M+H]//[M+Na]

R~140

R~140

R~135

R~140

CCS (A2)

O

O

HO

O ON

+ CH3

CH3CH3

POH

O

O

CH3

CH3

O

O

HO

O ON

+ CH3

CH3CH3

POH

O

O

CH3

CH3

O

O

HO

O ON

+ CH3

CH3CH3

POH

O

O

CH3

CH3

O

O

HO

O ON

+ CH3

CH3CH3

POH

O

O

CH3

CH3

O

O

HO

O ON

+ CH3

CH3CH3

POH

O

O

CH3

CH3

O

O

HO

O ON

+ CH3

CH3CH3

POH

O

O

CH3

CH3

Extraction blank 18 h starvedVegetative

Extraction blank

20% sucrose-fedStarved

Application example #1. Aedes aegypti raised on either a 20% sucrosesolution or water diet and dissected at four days. Ovaries submitted for lipidanalysis.

Application example #2. Lipid Profile Changes in nutrient-depletedDictyostelium discoideum at two different stages of cell cycle.

Rejectedbuckets

0 5 10 15 20 25 300

50

100

150

200

250

300

mSi

gma

Delta m/z (mDa)A B

A= average ofSPLASH lipids

B= 3xSD valuesof SPLASH lipids

IS-basedexclusion 0 10 20 30 40 50 60

CeramidesDAG

PCPEPG

PhosphosphingolipidsPIPS

Sphingoid basesTAG

Fed

Starved

Num

ber o

f lip

ids

iden

tifie

d

0 20 40 60 80 100

CeramidesDAG

Fatty amidesPCPEPG

PhosphosphingolipidsPI

SterolsTAG

Vegetative18 h starved

100

150

200

250

300

350

400

600

800

1000

0246

8101214

16

CC

S (A

2 )

m/z

RT (min)

PS

PA

Ceramides

DAG

Fatty amides

PC

PE

PG

Phosphosphingolipids

PI

Sterols

TAGLog2 fold change

-12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 60

1

2

3

4

5

-log1

0p-

valu

e

PS

PA

Ceramides

DAG

Fatty amides

PC

PE

PG

Phosphosphingolipids

PI

Sterols

TAG-3 -2 -1 0 1

0

1

2

3

4

5

6

Log2 fold change

-log1

0p-

valu

e

Num

ber o

f lip

ids

iden

tifie

d

Figure 5: TOP LEFT. Calibration curves obtained from normalized buckets. TOPRIGHT: ID parameters for deuterated lipids. BOTTOM: quantitation output ofmajor lipids from A. aegypti ovaries (LEFT) and D. discoideum cells (RIGHT).

TOTAL DAG

DG(18:1(9Z)/18:2(9Z,12Z)/0:0)[iso2]

DG(18:1(11E)/16:0/0:0)

DG(17:2(9Z,12Z)/19:0/0:0)[iso2]

TOTAL PC

PC(18:0/18:2(10Z,12Z))

PC(18:0/18:2(10Z,12Z))

PC(16:0/18:1(9Z))

PC(14:0/18:2(11Z,14Z))

PC(14:0/18:1(11Z))

TOTAL PE

PE(18:1(9Z)/16:1(9Z))

PE(18:0/18:1(9Z))

PE(16:1(9Z)/16:0)

PE(16:0/18:2(9Z,12Z))

PE(16:0/18:1(9Z))

PE(16:0/18:1(9Z))

TOTAL Phosphosphingolipids

SM(d18:1/18:0)

SM(d18:1/16:0)

PE-Cer(d16:2(4E,6E)/24:1(15Z))

PE-Cer(d14:2(4E,6E)/22:0)

PE-Cer(d14:1(4E)/22:0)

PE-Cer(d14:1(4E)/20:0)

TOTAL TAG

TG(16:0/16:1(9Z)/18:0)[iso6]

TG(14:1(9Z)/16:1(9Z)/18:1(9Z))[iso6]

TG(14:1(9Z)/16:1(9Z)/16:1(9Z))[iso3]

TG(14:0/16:1(9Z)/18:1(9Z))[iso6]

TG(12:0/16:1(9Z)/18:1(9Z))[iso6]

0.0 200.0 400.0 600.0 800.0 10k 100k

Lipid Content (pg/ovary)

Sugar-fed Starved

Inj. conc. (pg/ L)

0 100 200 300 400 500

Nor

m. I

nten

sity

0

10000

20000

30000

40000PC(16:0/18:1) [M+Na]+ PE(16:0/18:1) [M+Na]+ PG(18:0/18:1) [M+Na]+

PCY=77X-339R2=0.999

PEY=53X-924R2=0.960

PGY=22X-239R2=0.997

Deuterated lipid ID Class Mol. Form. RT [min]CCS (Å²) Main ion m/z

Δm/z [mDa] Fragmentation ions

PC(15:0/18:1)(d7) PC C41D7H73NO8P 9.4 284 [M+H]+ 753.611 -2.0 C5H13NO3P_166.0707(0.5929), HG_184.0787(100.0)

PC(18:1/0:0)(d7) PC C26H45D7NO7P 4.3 235 [M+H]+ 529.396 -2.9 HG_184.0797(100.0)

PI(15:0/18:1(d7) PI C42H72D7O13P 8.7 289 [M+H]+ 830.565 -11.7 HG_283.0203(100.0)

SM(d18:1/18:1(d9) Psphingolipid C42H72D7O13P 9.0 287 [M+H]+ 738.642 -4.9 Phosphocholine_184.0801(100.0)

TG(18:1/15:0/15:0)(d7) TAG C51H89D7O6 14.5 306 [M+NH4]+ 829.794 -4.5

14:0 C=O+_225.2294(24.6712), 17:1 C=O+_272.299(25.1047), G2_300.261(1.9702), M-18:1_523.4674(57.0314), M-15:0_570.5431(100.0)

ID buckets were generated by Bruker Metaboscape® 4.0 and annotated usingPremier Biosoft Simlipid® 6.04.

TOTAL PC

PC(O-14:1(1E)/0:0)

PC(O-6:0/O-6:0)

PC(18:2(2E,4E)/0:0)

PC(O-18:1(11Z)/0:0)

PC(O-16:0/4:0)

PC(18:2(2E,4E)/0:0)

TOTAL PE

PE(P-16:0/18:1(9Z))

PE(12:0/17:0)

PE(P-18:0/16:0)

PE(19:0/20:2(11Z,14Z))

PC(O-8:0/O-8:0)

PE(18:2(9Z,12Z)/0:0)

TOTAL Phosphosphingolipids

PI-Cer(t18:0/20:0(2OH))

PI-Cer(t18:0/18:0(2OH))

PI-Cer(t18:0/22:0(2OH))

PI-Cer(t18:0/16:0(2OH))

PI-Cer(t18:0/16:0)

SM(d18:1/16:0)

PE(16:0/0:0)

0.0 50.0 100.0 300.0 320.0

Vegetative 18 h Starved

Lipid Content (pg/cell)

Met

abo

lom

cis

and

Lip

ido

mic

s em

po

wer

ed b

y ti

msT

OF

9

Untargeted Metabolomics using Trapped Ion Mobility Spectrometry with Parallel Accumulation SErial Fragmentation (TIMS-PASEF)Karolina Sulek1*, Catherine G. Vasilopoulou2, Andreas-David Brunner2, Florian Meier2, Ulrike Schweiger-Hufnagel3, Aiko Barsch3 and Matthias Mann1,2

1 Novo Nordisk Foundation Center for Protein Research, University of Copenhagen, Denmark; 2 Max-Planck Institute of Biochemistry, Germany; 3 Bruker Daltonik GmbH, Germany* [email protected]

University of CopenhagenFaculty of Health and Medical Sciences Novo Nordisk Foundation Center for Protein Research

100

Reserpine[nM] TIMS off TIMS-PASEF

16 + +

1.6 + +

0.16 - +

0.016 - -

T-Rex4D

TIMS

PASEF

Figure 1. A. Instrument schematic of thetimsTOF Pro. B. TIMS operation from ionaccumulation to serial elution, where Erepresents electrical field and vg the gas flow[1]. C. Standard TIMS-MS/MS operation whereonly one precursor is selected from a singleTIMS scan.

Figure 2. Separation of two isomers D-lactose and D-maltose with m/z 365.1043 [M+Na]+ injected in timsTOF Proinstrument in ramp time 350 ms and accumulation time 100ms.

Figure 3. Correlation betweenexperimental (CCSexp) and literature(CCSlit) collisional cross section valuesfor different classes of metabolitesand lipids in human plasma extracts[2,3]. LPC–Lysophospholipids, PC-Phospholipids, SM – Sphingomyelins,Hormone-Testosterone.

Table 1. Absolute CCS values (A2) and % deviation between experimentaland literature data for metabolites and lipids from human plasma extracts.

Table 2. Reserpine spiked into metabolicextracts of human plasma at differentconcentrations.

.

Plus (+) indicates that an MS/MS spectrum forthe reserpine precursor ion was acquired. TIMS-PASEF outperforms the standard acquisition interms of sensitivity due to the increasedsequencing speed.

Figure 4. Principle of PASEF operation. Rapid switching of the quadrupole mass position enables selection of multiple precursorsat different m/z from a single TIMS scan. This multiplies the MSMS acquisition rate and allows to acquire hundreds of MSMS persecond in LC-MS experiments, without decreasing the sensitivity [4,5].

B.

A.

C.

A.

B.

D.

C.

Figure 5. A. Number of MSMS scans and B. unique precursors in TIMS-MSMSand TIMS-PASEF mode in human plasma metabolic extracts in a 22min LC run.C. High sequencing speed of PASEF can be used to sequence low-abundanceprecursors repeatedly. D. Number of precursors per PASEF scan.

Figure 7. Data processing overview using a beta version of MetaboScape 4.0. Rawfiles from timsTOF Pro were used by the T-Rex 4D processing algorithm for four-dimensional feature detection, peak alignment and extraction of CCS values.Identification of metabolites is based on m/z, RT and MS/MS spectral libraries(MetaboBase, HMDB, LipidBlast). Ion mobility values of identified compounds canbe read out to generate CCS libraries to improve specificity for future analyses.

Figure 6. LC-TIMS-PASEF data acquisition of human plasma metabolic extracts in 22 min LC run. Raw file view shows multi-dimensionalseparation of molecules according to their retention time, mass to charge ratio, mobility,, and PASEF fragmentation pattern.

Figure 8. A. Number ofdetected features byMetaboScape 4.0 fromhuman plasma extract withassigned MSMS spectra inPASEF mode vs. MS/MSscan. B. T-Rex algorithmresults of CCS values inmixture of referencemetabolites and lipids infour technical replicates;numbers indicate thecoefficient of variation.

= 1.4%

A.

B.

[1] Ridgeway ME, Lubeck M, Jordens J, Mann M, Park MA (2018) Int. J. MS; 425, 22-35.[2] Chouinard DC, Beekman CR, Kemperman RHJ, King HM, Yost RA (2017) J Am Soc Mass Spectrom.; 20, 31-39.[3] Paglia G; Williams AP; Richardson K; Olivos, HJ; Thompson, JW; Menikarachchi, L; Lai S; Walsh C; Moseley A.; Plumb RS; Grant DF; Palsson BO; Langridge J; Geromanos S; Astarita G (2015) Anal Chem.; 87(2), 1137-44.[4] Meier F, Beck S, Grassi N, Lubeck M, Park MA, Raether O, Mann M (2015) J Proteome Res.; 14(12), 5378-87.[5] Meier F, Brunner A-D, Koch S, Koch H, Lubeck M, Krause M, Goedecke N, Decker J, Kosinski T, Park MA, Hoerning O, Cox J, Raether O, Mann M (2018) bioRxiv; doi: https://doi.org/10.1101/336743

Name CCSexp CCSlit Name CCSexp CCSlit

Testosterone 176 175 0.57 PC 32:1 281 287 2.09

LPC 14:0 223 227 1.76 PC 34:2 292 293 0.34

LPC 16:0 237 236 0.42 PC 36:2 297 301 1.33

LPC 16:1 226 229 1.31 PC 38:4 299 305 1.97

LPC 18:1 239 240 0.42 PC 38:5 298 302 1.32

LPC 20:3 240 249 4.02 PC 40:6 301 309 2.59

LPC 20:4 237 242 2.07 SM (d18:1/16:0) 287 291 1.37

ms

11.3 fold

Metabolomcis and Lipidomics empowered by timsTOF

10



Applying Trapped Ion Mobility separation (TIMS) in combination with Parallel Accumulation Serial Fragmentation (PASEF) for analysis of lipidomics samples

Sebastian Götz1; Sven W Meyer1; Ulrike Schweiger-Hufnagel1; Aiko Barsch1; NingombamSanjib Meitei21Bruker Daltonik GmbH, Fahrenheitstraße 4, 28359 Bremen, Germany 2 PREMIER Biosoft, Indore, India

IntroductionAiming at a deeper understanding of biochemistry, the analysis of lipid species is an important area. Untargeted lipidomics workflows target the profiling of changes in the lipidome in order to discover relevant lipids as potential biomarkers. This approach relies on robustly identifying a large number of lipids and statistically evaluating their relative abundances. We present an improved lipidomics identification workflow that is based a combination of LC-MS and TIMS separation together with a very fast data dependent MS/MS fragmentation (PASEF)1. The comprehensive data sets are processed and statistically evaluated with MetaboScape software in combination with SimLipidsoftware for identification of the lipids. Serum samples of 4 different species were evaluated (human, pig, chicken, bovine).

Serum samples were extracted using a liquid-liquid extraction method (methanol/methyl tert-butyl ether). LC separation was performed using a Bruker Elute UHPLC system (32 min gradient program)2. MS data acquisition was performed in Parallel Accumulation Serial Fragmentation mode (PASEF) in combination with TIMS separation on a Bruker timsTOF PRO instrument (see Figure 1).

The instrument was set to cover a broad mobility range in TIMS mode while acquiring MS/MS fragmentation spectra in Parallel Accumulation Serial Fragmentation mode (PASEF, see Figure 2) during the LC gradient. The resulting data sets were processed in a novel MetaboScape 4.0 software (Bruker) version using a 4-dimensional feature finder designed to include the mobility separation dimension. Features were annotated using SimLipid6.04 software (PREMIER Biosoft).

ResultsMethodsFeasibility of the TIMS system was first checked using a targeted infusion method with highest mobility separation settings. Two close isobaric lipids (18:0-18:1 PE and C16-18:1 PC) were mixed and both the protonated and sodiatedions could be well separated in the mobility dimension with TIMS resolution up to 180 (Figures 3a) and 1b)). A broader untargeted TIMS-MS method was then developed and applied to a 32 minute LC gradient separation2.

Extracted serum samples of the different species were analyzed using this method. To accommodate for the additional dimension of information in LC-TIMS-MS data a dedicated 4-dimensionsal feature finder algorithm was developed and implemented into MetaboScape software. The algorithm (T-ReX 4D) enables robust detection of features including recursive extraction and automatically assigns available PASEF MSMS spectra to the found MS1 features. In this evaluation only features were accepted that could be found in all 3 replicates of an analyzed serum sample. All features with associated MSMS spectra were then exported to SimLipid software for lipid identification.

SimLipid allows for the identification of multiple lipids in a single MSMS spectrum. These chimeric spectra were still occasionally observed, but only the single best lipid ID per MSMS spectrum was used for the presented results. Although SimLipid implements a proprietary algorithm that allows for identification of fatty acid chain length due to fragmentation patterns, this list was further filtered and condensed to avoid duplicates and over-specification by using short names (e.g. PC(32:1) instead of e.g. PC(16:0_16:1)). Table 1 shows number of identified lipids for the different evaluation methods. Most abundant main classes of identified lipids were glycerophosphocholines(PC), triacylglycerols (TAG), phosphosphingolipids and glycerophosphoethanolamines (PE). Relative abundances of these classes per analyzed species are shown in Figure 4.

Figure 5 shows MetaboScape results after processing all species sample replicates in a single feature list (bucket table). The software allows for running multiple statistical analyses (Principal Component Analysis used here). The scores and loadings plots show a clear separation of the groups with a high degree of reproducibility regarding the replicates. Automatically generated collisional cross section values (CCS) are available for each feature facilitating alternative library search workflows with increased identification confidence.

Conclusions• Targeted approach allows TIMS

separation of (close) isobaric lipids with resolution of 180

• Newly developed 4-dimensional feature finder (T-ReX 4D) in MetaboScape enables thorough and reproducible extraction of all relevant features including collisional cross sections and corresponding MSMS Information

• SimLipid and MetaboScape were integrated in a streamlined workflow for enabling LC-TIMS-PASEF based lipidomics research)

timsTOF / PASEF

References(1) Meier et al.; J. Proteome Res. 2015, 14:5378-5387(2) Wörmer et al.; Org Geochem 2013, 59:10-21

2

1

18_0-18_1 PE + C16-18_1 PC-001.d: EIM 746.6062±0.05 +All MS, 0.0-0.8 min, Smoothed (0.001,1,SG)

3

4

18_0-18_1 PE + C16-18_1 PC-001.d: EIM 768.5521±0.05 +All MS, 0.0-0.8 min, Smoothed (0.001,1,SG)0.0

0.2

0.4

0.6

0.8

1.05x10

Intens.

0

1

2

3

4

5

64x10

1.38 1.40 1.42 1.44 1.46 Mobility, 1/K

Human Pig Chicken Bovine Lipid IDs (common in all replicates) 442 234 330 270 Unique lipid IDs (short name) 340 208 255 232

Figure 3a) Heatmap of close isobaric PC/PE lipid pair separated in mobility dimension (targeted infusion method)

[M+H]+

[M+Na]+

Figure 3b) Mobiliograms for protonated and sodiated ions of PC/PE lipid pair (targeted infusion method)

resolution115 / 135

resolution178 / 187

Figure 4) Relative abundance of lipid main classes per species

Table 1) Number of identified lipids in different species

[M+H]+

[M+Na]+

Figure 5) Serum samples of 4 different species processed in MetaboScapesoftware

Figure 1. Ion optics of the timsTOF Pro instrument including a dual TIMS analyzer and QTOF mass spectrometer.

Figure 2. Illustration of the PASEF method in comparison with the standard TIMS MS/MS

Metabolomcis and Lipidomics empowered by timsTOF

11

Generation of a Collisional Cross Section Library for Plant Metabolomics using Trapped Ion Mobility Spectrometry (TIMS)

Conclusion• We are creating a CCS library emphasized on plant natural products which will improve identification

confidence in metabolomics.• Isomeric and structurally similar compounds can be differentiated by TIMS even when they can not be

differentiated on mass alone.• TIMS technology and PASEF is promising for improving our depth of coverage into the metabolome

ACKNOWLEDGMENTS; This work was supported by: The University of Missouri, and a collaborative effort with Bruker Daltonics (Aiko Barsch and Sven Meyer)Structures from National Center for Biotechnology Information. PubChem Compound Database

Results

Compound Name MW Ion CCS average

(Å²) CCS

Stdev CCS %RSD

2'-Hydroxyflavone 238.25 [M-H]- 149.55 0.07810 0.0522%



3-Hydroxyflavone 238.25 [M-H]- 151.79 0.18502 0.1218%




6-Hydroxyflavanone 240.07 [M-H]- 156.51 0.15133 0.0966%

A

B

C

We have focused on two groups of specialized metabolites: flavonoids andsaponins. These compounds can be hydroxylated, methylated, and glycosylated atvarious positions. We have recorded CCS for over 150 unique compounds. TheCCS average, standard deviation, and relative standard deviation (%RSD) werecalculated for each compound analyzed in triplicate. An acceptable %RSD is


Data Analysis Workflow

The bottleneck for non-targeted profiling approaches is the annotation of relevant compounds, which is needed for the biologicalinterpretation of the results. In fact, studies publishing differential compounds rarely reach annotation rates above 20 % [1]. Publiclyavailable databases and the development of customized tools for metabolite annotation facilitate the batch annotation of previouslyknown and unknown metabolites of biological relevance.

References:[1] Blaženović, I.; Kind, T.; Ji, J.; Fiehn, O. Metabolites 2018, 8.[2] Ruttkies, C.; Schymanski, E. L.; Wolf, S.; Hollender, J.; Neumann, S. J Cheminform 2016, 8, 3.[3] Treutler, H.; Tsugawa, H.; Porzel, A.; Gorzolka, K.; Tissier, A.; Neumann, S.; Balcke, G. U.

Analytical Chemistry 2016, 88, 8082-8090.

Department of Stress and Developmental Biology

Untargeted Metabolomics in Plant BiochemistryStefanie Döll, Hendrik Treutler, Steffen Neumann, Dierk Scheel

Experimental Set up Quality Assurance

• Injection blanks

• Extraction blanks

• External standards

• Internal standards

• Mixed sample (QC)

• Randomization

0

2000

4000

6000

8000

10000

12000

1840

818

412

1841

618

420

1842

418

428

1843

218

436

1844

018

444

1844

818

452

1845

618

460

1846

418

468

1847

218

476

1848

018

484

1848

818

492

1849

618

500

1850

418

508

1851

2

injection number

IAA-Valin

Processing: Feature extraction

neg-E571_cerk1-1 E2_K-4_Wurzel_1-C,1_01_16372.d: BPC -All MS

neg-E571_cerk1-1 E2_C8-1_Wurze_1-B,4_01_16367.d: BPC -All MS



neg-E571_cerk1-1 E2_C8-4_Wurze_1-C,2_01_16373.d: BPC -All MS

neg-E571_cerk1-1 E2_C8-5_Wurze_1-C,3_01_16375.d: BPC -All MS

neg-E571_cerk1-1 E2_K-1_Wurzel_1-B,3_01_16366.d: BPC -All MS



0.0

0.5

6x10Intens.

0.0

0.5

6x10Intens.

0.0

0.5

6x10Intens.

0.0

0.5

6x10Intens.

0.0

0.5

6x10Intens.

0.0

0.5

6x10Intens.

024

5x10Intens.

0.0

0.5

6x10Intens.

0.0

0.5

6x10Intens.

0.0 2.5 5.0 7.5 10.0 12.5 15.0 Time [min]

1. Massrecalibration

2. Retention time alignment

3. Region complete featureextraction- grouping of isotopes,

fragments, adducts andcharge states

- search for „missing“ peaks

higher in „control“

higher in „treatment“

volcano plot for t-test

Feature extraction, Alignment and Peak Grouping Univariate + Multivariate Statistics

Compound identification In-silico fragmentation

LC-Separation

Extraction MS-Analysis

UV +Fluorescence

Detection

Data Preprocessing

Data Analysis

ExperimentalDesign

Sampling & Preparation

Sample Preparation LC-MS

Data Analysis

MSMS-Analysis

• Calibration• Peak picking• Alignment• Filling in missing

data

• Replicates(biol. + techn.)

• Controls• Data analysis

strategy

Selection of • Gradient• Column• Phase

Introduction

We support two workflows:• Quantification of a predefined set of

metabolites (analyte list, metabolite class)

Study a hypothesis quantitative descriptionTa

rget

ed: • Filter out and identify substances relevant to

the investigated subject

Unt

arge

ted:

Find metabolites characteristic for, e.g.:mutant vs. wild type treatment vs. control healthy vs. infected

+

• Discovery of diverging metabolite families usingthe web app MetFamily [3]

• Two orthogonal analyses• Principal component analysis (PCA):

differentially accumulating compounds from MS data• Hierarchical cluster analysis (HCA):

Metabolite families from MS2 data• HCA+PCA: Discriminating metabolite families!• Freely available and open source:

https://msbi.ipb-halle.de/MetFamily/

Acyl sugarsSesquiterpeneglucosides

https://msbi.ipb-halle.de/MetFragBeta/

LC-ESI-HR-QTOF-MS

Electrosprayionisation (ESI)

Size exclusion+ focussing

Time-of-Flight measurement

Optional: Fragmentation(MSMS)

LC separation: C18, Water/ACN+FA

Clustering: unscaled euclidean distance

WT

WT

Mutant

Mutant

control

treatment

PCA 3D

Metabolite classification

Browser based fragmentation ofcompounds from several databases [2]

Spe

ctra

ldat

aba

sese

arch

In-h

ouse

libr

arie

s

High resolution LC-QTOF-MS based Metabolomics

13


Metabolomics Profiling and Identification of Cell Culture Media

Xuejun Peng1; Guillaume Tremintin1; Anjali Alving2; Nikolas Kessler2; Heiko Neuweeger3, Aiko Barsch3

1Bruker Daltonics, 61 Daggett Drive, San Jose CA USA 95134; 2Bruker Daltonics, 40 Manning Road, Billerica MA USA 01821; 3Bruker Daltonik GmbH, Fahrenheitstraße 4, Bremen Germany 28359

IntroductionMetabolomics Profiling and Identification of cell culture media is of increased interest. Comprehensively profiling up to 95 compounds in culture media including amino acids, monosaccharides, vitamins, nucleic acids, antibiotics, and other primary metabolites was conducted by targeted and non-targeted metabolomics simultaneously, aiming to establish a media profile that can be correlated to cell growth and quality.

The comprehensive data sets were processed and statistically evaluated with MetaboScape 4.0 program (Figure 2) in combination with DataAnalysis software for profiling and identification of the components in cell culture media.

Results and Discussions

Data Analysis

Profiling of Cell Culture Media

It becomes increasing important to profile the ingredients of starting cell culture media, nutrients depletion or change during cell culture process, and the formation new metabolites. The nutritional components in cell culture media promote cell growth, maintain cell function and status, and influence the final protein production, and its productivity yield and quality. About 95 commonly used and important biological nutrients in cell culture media, including amino acids, monosaccharides, vitamins, nucleic acids, antibiotics and others, were screened against purchased different commercial cell culture media. Its LC-MS separation chromatogram (Figure 3) could be used as fingerprint to monitor batch-to-batch variation and quality of cell culture media.

Selection of Cell Culture Media

Because cell culture media greatly influences cell growth process, to find an appropriate starting cell culture medium is critical to cell growth performance. For example the amino acid composition and concentration in the media is critical since amino acids make up the primary structure of a therapeutic protein in cell culture, and support maximal cell growth and sustain cell viability at increasing cell densities. It was observed (Figure 4) that glucosamine concentration in NCTC-109, HAM-F10, HAM-F12 and McCoy-5A from same vendor are different.

Vendor Evaluation of Cell Culture Media

It was noticed (Figure 5) that Leucine in HAM’s F12 cell culture media from different vendors show different intensity level. In order to maintain consistent cell growth, it is essential to establish

a screening analytical method and monitor the nutritional components in starting cell culture media, even it was obtained from same vendor in order to avoid lot to lot variability and maintain reproducible cell growth process.

Spent Media Analysis

Spent media was prepared by centrifuging and separating the supernatant from cells, followed by preparative IEX and SEC, then sterile filtered by 0.2µm. Based on the profiling patterns of different spent media [Figure 3 (b1, b2)], we could quickly see the relationship of required nutrients and the metabolism process, and further optimize the cell culture bioprocess conditions according to the protein production and its quality. The intensity differenc

Metabolomics at Bruker · prediction model (see Figure 3). Figure 4 highlights the workflow for...

Documents

Transcript of Metabolomics at Bruker · prediction model (see Figure 3). Figure 4 highlights the workflow for...