Mobilising Ion Mobility Mass Spectrometry for Metabolomics

Eleanor Sinclair1,2, Katherine A. Hollywood2, Cunyu Yan2, Richard Blankley3, Rainer Breitling2,

Perdita Barran1,2

1Michael Barber Centre for Collaborative Mass Spectrometry, Manchester Institute of Biotechnology,

School of Chemistry, The University of Manchester, Princess Street, Manchester, M1 7DN, UK

2BBSRC/EPSRC Synthetic Biology Research Centre for Fine and Speciality Chemicals (SYNBIOCHEM), Manchester Institute of Biotechnology, The University of Manchester, Manchester, U.K

3Agilent Technologies, Life Sciences & Chemical Analysis Group, Cheadle Royal Business Park, Cheshire, UK, SK8 3GR

Abstract

Chromatography based mass spectrometry approaches (xC-MS) are commonly used in

untargeted metabolomics, providing retention time, m/z values and metabolite specific-

fragments all of which are used to identify and validate an unknown analyte. Ion mobility-

mass spectrometry (IM-MS) is emerging as an enhancement to classic xC-MS strategies, by

offering additional ion separation as well as collision cross section (CCS) determination. In

order to apply such an approach to a metabolomics workflow, verified data from metabolite

standards is necessary. In this work we present experimental DTCCSN2 values for a range of

metabolites in positive and negative ionisation modes using drift tube-ion mobility-mass

spectrometry (DT-IM-MS) with nitrogen as the buffer gas. The value of DTCCSN2

measurements for application in metabolite identification relies on a robust technique that

acquires measurements of high reproducibility. We report that 86 % of the metabolites

measured in replicate have a relative standard deviation lower than 0.2 %. Examples of

metabolites with near identical mass are demonstrated to be separated by ion mobility with

over 4 % difference in DTCCSN2 values. We conclude that the integration of ion mobility into

1

current LC-MS workflows can aid in small molecule identification for both targeted and

untargeted metabolite screening.

Introduction

Metabolites, typically organic molecules with a mass no greater than 1500 Da, constitute

the entire complement of metabolic processes within an organism and they cover a vast

range of diverse chemical species, including; amino acids, carbohydrates, lipids and vitamins

amongst many others.(1) The diversity of this domain naturally prompts the question of

how best to analyse the components. Due to the low sample volumes required and low

limits of detection, mass spectrometry has a natural disposition for these types of analyses,

giving it an advantage over alternative popular techniques, such as Nuclear Magnetic

Resonance (NMR), whose success depends on a higher sample concentration and volume.

(2)

Due to the complexity of metabolomics samples, mass spectrometry detection alone is

normally insufficient, prior to mass detection a chromatographic separation is typically

employed to aid metabolite detection and identification.(1–4) Liquid chromatography (LC) is

arguably the most routine technique adopted and the most applicable to a wide-range of

samples; however the technique is not without its drawbacks. The hydrophobicity of the

sample in question impacts on the LC separation approach chosen, and so a priori

knowledge of the analyte polarity is highly preferable, highlighting a limitation for

untargeted metabolomics analysis.(3,4) The lack of reproducibility with reference to

variance in sample matrices, sample loading and even slight variability between column

manufacturers can lead to a shift in retention times (RT), and hence complicates database

2

matching for metabolite identification.(5) Despite such so-called disadvantages, LC-MS is

routinely employed in metabolomics investigations, where its separation capability

increases the dynamic range for detection of metabolites.

Ion mobility (IM) is a gas phase separation technique which separates ions based on their

size, shape and charge. In the presence of a buffer gas, typically nitrogen (N2) or helium (He),

ions are released into a drift tube across which a weak electric field is applied. The

characteristic time an ion takes to traverse this chamber is its drift time, which for

electrostatic drift fields can be converted to a collisional cross section (CCS) via the Mason-

Schampp equation as long as the temperature and pressure of the pure gas are known along

with the mass and charge of the ion. The CCS value is a composite parameter which is

influenced by the rotationally averaged three-dimensional conformation of a given ion and

how that interacts with the neutral buffer gas. (5–8). For a single conformer, drifting in a

pure neutral gas the CCS should be an inherent and reproducible property. Hyphenating ion

mobility to an LC time-of-flight (ToF) mass spectrometer has distinct analytical advantages,

with a timescale of separation (milliseconds) that nestles between LC (seconds to minutes)

and MS (microseconds).(6,9) In addition to adding an extra dimension of information to

typical analyses, IM also extends the degree of separation in samples, enabling metabolites

to be separated from chemical noise and matrix background.

Several studies published in recent years convey the benefits of such an approach in small

molecule identification; investigating the distribution of CCS values within a molecular class

and the creation of databases (5,7,10,11), improvements to the mobility measurement

(12,13) and exemplar systems such as in the IM separation of protomers (14).

3

Experimentally measured CCS values can be compared with computationally derived

theoretical CCS values obtained using for example molecular modelling (5,13,15) or machine

learning algorithms (16).

The objective of this work is to determine if IM-MS could be used to enhance mass

spectrometry workflows for small molecule identification. The aforementioned limitations

of chromatography with respect to stability and inter-instrument comparability of RT leads

to difficulties in assigning confident identifications to small molecules. Ion mobility shows

great potential in enhancing a method for robust analytical characterisation of the

metabolome.

Experimental Section

Chemicals: All chemical standards were purchased from SigmaAldrich, and a comprehensive

list of them can be found in Supplementary Information Tables S1 and S2. MS-calibration

mixture (ESI low concentration tune mix) was obtained from Agilent Technologies (Santa

Clara, CA), containing 12 compounds including 9 hexakis-(fluoroalkoxy)-phosphazenes,

ammonium trifluoroacetate, betaine and 2,4,6-tri(heptafluoroproypyl)-1,3,5-triazine.

Acetonitrile purchased from SigmaAldrich, methanol from Thermo Fisher Scientific Inc.

(NYSE, TMO) and water obtained from a 18.2 M Ω cm at 25°C Milli-Q system (Merck

Millipore Corporation, Billerica, MA). All chemicals and solvents used were of analytical

grade or above in purity.

Instrument: UHPLC systems (1290, Agilent Technologies) were utilised primarily as a

method of infusion for the standards, and in assessing the success of ion mobility

implementation into an LC-MS workflow, however development of a robust separation

4

method was not a focus. A Waters XBridge Amide 3.5 µm (4.6 x 100 mm) column was

employed for separation, and in both positive and negative ionisation modes mobile phase

A was: pH=9.0, 95 % water (vol/vol), 5 % acetonitrile (vol/vol), 20 mM ammonium

hydroxide, 20 mM ammonium acetate, mobile phase B was: 100 % acetonitrile.(17) The

elution gradient as follows was used; 0 min 8 5% B, 7 min 65 % B, 12 min 35 % B, 13 min 10

% B, 15 min 10 % B, 15.5 min 85 % B, 21 min 85 % B. All experiments used a 600 µL flow

rate, 55 °C column temperature and an injection volume between 3-8 μL was selected,

which was determined dependent on the sample concentration and the response of the

instrument.

Experimental: DTCCSN2 values1(18,19) were determined utilising two identical DT-IM-MS

instruments (6560, Agilent Technologies), coupled to UHPLC systems using nitrogen buffer

gas. In brief gaseous ions are generated within a dual Agilent jet stream (AJS) ESI source,

after which they are transferred and focused through front ion funnels. Prior to separation

they are trapped, accumulated and subsequently a ‘packet’ of ions is released into ~78 cm

long drift tube. This chamber is comprised of a series of ring electrodes at which direct

current (DC) voltage is applied, changing linearly through the drift tube. Following ion

mobility separation the ions are refocused in the rear funnel before they pass through a

quadrupole, collision cell and finally are detected by a time-of-flight mass analyser.(8,12,20)

Both instruments were tuned using the Agilent MassHunter Acquisition (B.07.00) Auto Tune

feature in the 50-1700 m/z range, and a list of any manually altered instrumental

parameters can be found in the Supplementary Information Table S3.

Analysis: DTCCSN2 values were calculated using Agilent MassHunter IM-MS Browser

(B.07.01). Experimentally obtained drift time profiles have been converted to collision cross

1 The syntax DTCCSN2 has been described by us previously to refer to the type of ion mobility (here DT = drift tube) and the gas in which the measurement is made.

5

section distributions (CCSD). Each CCSD considers any inter-experiment variance in

temperature and pressure and represents an average over three replicates acquired for

each respective metabolite.

Single Field DTCCSN2 determination using DT-IM-MS:

DT-IM-MS allows the experimental drift time of an ion to be directly correlated to its

respective DTCCSN2 value, (see Supporting Information for further details). The ‘stepped field’

methodology requires an experimental parameter to be varied in the duration of the

experiment, most commonly the electric field. Whilst this is readily coupled to direct

infusion sample injections, it is unfeasible to integrate it in to the chromatographic

timescales required for the separation of complex mixtures.

An alternate method for DTCCSN2 determination utilises a calibration approach, facilitating its

addition within established LC-MS workflows. The method uses a series of calibration ions

each of which have well characterised DTCCSN2 values from stepped field experimentation

and these values then form the basis of the single field measurement. Constructing a plot of

experimentally measured drift times (tD) against standardised DTCCSN2 (Ω) of the calibrant

ions allows a linear regression to be fitted. The two unknown calibration terms from the

equation below (1), β and tfix, are determined from the slope and intercept of this regression

line.

tD=βz [ mimi+mB ]

1 /2

Ω+t fix [1]

6

Combining these calibration terms with the experimentally measured terms; drift time (tD),

mass (m) and charge state (z), enables an unknown CCS (Ω) to be calculated with the

rearrangement of Eqn. 1. For accurate DTCCSN2 determination both the tune mix calibrant

solution and the sample in question should be performed at an identical electric field, drift

tube pressure, drift gas temperature and under the same experimentally applied

parameters. This single field calibration method for DTCCSN2 determination has been applied

for the measurements reported in this paper.

Results and Discussion

DTCCSN2 values have been measured for 50 metabolites in negative ionisation mode and 35

metabolites in positive ionisation mode; the most dominant species presenting as

deprotonated and protonated ions respectively, the values of which are reported in the

Supplementary Information Tables S1 and S2. Figure 1 displays a plot of these CCS values

against the measured m/z values and the corresponding error associated with each value is

too small to allow for visualisation on the axis scales. A table of standard deviations (S.D.)

can again be found in the Supplementary Information Tables S1 and S2, and subsequently

represented in Figure 3. Metabolites covering a mass range of 110 Da to 525 Da were

analysed in both ionisation modes and were found to exhibit a DTCCSN2 range spanning 115

Å2 to 200 Å2 across both polarities. Both plots show good correlation between mass and

collision cross section with R2 values of 0.9565 and 0.8574 for positive and negative mode

respectively. As might be expected given the broad chemical diversity of these example

compounds the ‘trend’ is not a strictly linear relationship thus affirming the benefit of

coupling IM to MS. The difference in R2 values between the two ionisation modes can in part

7

be accounted for by the difference in the number of species analysed in each polarity. If we

only consider the common metabolites measured between polarities, the R2 values are

0.9363 and 0.9060 for positive and negative modes respectively i.e. these are not the same.

Whilst the observation that CCS increases as a function of mass is intuitively expected and

has been noted previously, (8,11) we highlight here that enhanced separation between an

IM-MS and MS experiment is also dependent on ionisation.

Figure 1: Scatter plots of collision cross-section values against mass-to-charge ratio, linear regression fitted to each data series displayed as red dashed lines, (A) 35 metabolites in positive ionisation mode, protonated [M+H]+ species, R2=0.9584 . (B) 50 metabolites in negative ionisation mode, deprotonated [M-H]- species, R2=0.8574

The vast majority of metabolites are below 1000 Da in mass, organic in nature, and their

structure is not dictated by non-covalent inter/intra-molecular interactions, often limiting

their conformational diversity when compared to larger biomolecules and macromolecular

complexes.(21–23) However, there are still large areas of diversity displayed both in their

structure and in their chemical functionality. The primary basis of ion mobility is not

founded on m/z separation but on separations brought about by collisions with the buffer

gas present. Ions are discriminated dependent on the so-called conformational space they

8

(B)(A)

occupy, and to a degree, on interactions between the specified buffer gas and any

functional groups present.(14,20)

Figure 2: Collision cross section distributions highlighting the drift time separation of close in mass species . (A) Two metabolites, homocystine (left) and inosine (right), differ by 0.0256 Da (0.01 %) in monoisotopic mass and by 4.4 % in DTCCSN2 values. (B) Two flavonoids, quercetin (left) and hesperetin (right), differ in monoisotopic mass by 0.0364 Da (0.01 %) and display a 6.6 % difference in DTCCSN2 values;

Figure 2 shows two examples of small mass differences between molecules that link to a

comparatively large differentiation based upon their DTCCSN2 values. Figure 2(A) highlights

the IM separation of homocystine and inosine. These two molecules do not share a common

molecular class, or similarities in their structural composition; however, they do share a

similar monoisotopic mass exhibiting a difference of 0.0256 Da. This 0.01 % mass difference

is associated to an unexpectedly large mobility separation of 4.4 % in measured DTCCSN2

values.

Extending from this, Figure 2(B) again demonstrates a system that exhibits a large DTCCSN2

separation in comparison to the respective masses of the molecules. Quercetin and

hesperetin both belong to the flavonoid molecular class, and thus both share the same

9

(B)(A)

polyphenolic structure. The 0.0364 Da difference in monoisotopic mass arises from minor

structural variances in the functional groups present, which results in a 6.6 % difference in

DTCCSN2 values; and a fully resolved ion mobility separation can be seen.

Figure 3: Bar charts displaying the spread of relative standard deviation (%) of DTCCSN2 values taken from

replicate experimental acquisitions on one DT-IM-MS instrument, each horizontal bar represents a

metabolite measured in the study; (A) in positive ionisation mode, corresponding to 86 % of measured

metabolites under 0.2 % R.S.D. and (B) negative ionisation mode with 90 % of metabolites studied under 0.2 %

R.S.D.

Due to the thousands of compounds that encompass the metabolomes of various

organisms, all of which are found within a relatively narrow mass range, there is a

requirement for a high level of accuracy and reproducibility in the analytical measurements

made. Standard deviation of CCS values within 2 % (5,24) has been previously quoted as an

accepted experimental error, however this guideline limits the potential for CCS

measurements to be useful as a molecular identifier. Figure 3 demonstrates the relative

standard deviation (R.S.D.) obtained over replicate measurements for each metabolite

measured in this study, in positive (Fig 3A) and negative (Fig 3B) ionisation modes. Of the

metabolites measured 86 % and 90 % were found to have an R.S.D. less than 0.2 % in

10

(B)(A)

positive and negative ionisation modes. This corresponds to errors that are an order of

magnitude lower than previously accepted and has therefore been used as a comparator

value in Figure 3. The variations in S.D. do not seem to correlate to m/z and so we propose

that this is linked to the structure and the elemental composition of each metabolite.

Figure 4: Collision cross section distributions of four metabolites, representing an average from triplicate

measurements, obtained from two identical DT-IM-MS instruments and overlaid, IM-MS system 1 (blue)

overlapped with IM-MS system 2 (red). (A) phenylalanine [M-H]-, (B) folic acid [M+H]+, (C) guanosine [M+H]+,

(D) riboflavin [M+H]+. The corresponding S.D. is labelled aside each peak in both Å 2 and as a percentage value

which have been generated from the standard deviation calculation of averaged triplicate measurements from

each instrument.

The reproducibility of DTCCSN2 measurements acquired on one IM-MS system has been

established (Figure 3); in addition to this, the question of reproducibility across IM-MS

systems has been probed. Figure 4 begins to answer this, displaying exemplar overlaid

11

DTCCSN2 distributions acquired on two identical DT-IM-MS instruments (6560, Agilent

Technologies), for four of the metabolites studied within this project. Each CCSD is a

representation of triplicate measurements taken on each IM-MS instrument; S.D. values

have been calculated between the replicate sets acquired on each respective instrument.

Subsequently standard deviations between the averaged DTCCSN2 values obtained on each

instrument are displayed beside the distributions in Figure 4. The near-perfect overlap of

the CCSD peaks in the figure arises from the similarities between the distributions acquired

on each IM-MS system, both with respect to the apex and breadth of each CCSD peak. The

values assigned as standard deviation within Fig 4a-4d reflect the difference in DTCCSN2 value

obtained from peak apex and no statistical analysis is reported regarding peak width.

Further metabolites will need to be studied for any conclusions to be drawn on the

repeatability of DTCCSN2 measurements across instruments; however these preliminary

experiments exhibit promising results.

Conclusions

In this study we successfully integrated ion mobility into an existing LC-MS workflow on a

DT-IM-MS instrument. The focus of the results was directed to using DTCCSN2 values as a

molecular identifier in small molecule analytics. A correlation between mass and CCS has

been shown alongside examples of large ion mobility separation for molecules with small

mass differences, a 0.01 % mass difference being associated to DTCCSN2 separations between

4.4 % and 6.6 %, both with respect to similar and contrasting chemical structure. CCS

determination has been performed at a single electric field using a set of calibration

standards which yielded high reproducibility. Over 85 % of molecules exhibited relative

standard deviations beneath 0.2 %, measured over replicate measurements in both positive

12

and negative ionisation modes, for both intra- and inter-instrument replicates. The high

level of reproducibility and the extra data in the form of CCS values and distributions occur

without any increase in the analytical acquisition time. Further to this, the enhanced

separation of IM has the potential to reduce the time taken for each experiment by the

reduction or omission of the chromatographic component.

Acknowledgements

This work was supported by the BBSRC in its award BB/M017702/1, and we acknowledge

the entire SYNBIOCHEM team for their ongoing contributions in support of this work.

Agilent Technologies Inc. are thanked for a University Relations Award which along with an

EPSRC DTA grant to the School of Chemistry has funded the PhD. of ES.

References

1. Dunn WB, Ellis DI. Metabolomics: Current analytical platforms and methodologies. TrAC - Trends Anal Chem. 2005;24(4):285–94.

2. Dunn WB, Broadhurst DI, Atherton HJ, Goodacre R, Griffin JL. Systems level studies of mammalian metabolomes: the roles of mass spectrometry and nuclear magnetic resonance spectroscopy. Chem Soc Rev. 2011;40(1):387–426.

3. Theodoridis GA, Gika HG, Want EJ, Wilson ID. Liquid chromatography-mass spectrometry based global metabolite profiling: A review. Anal Chim Acta. 2012;711:7–16.

4. Dettmer K, Aronov PA, Hammock BD. Mass spectrometry-based metabolomics. Mass Spectrom Rev. 2007;26(1):51–78.

5. Paglia G, Williams JP, Menikarachchi L, Thompson JW, Tyldesley-Worster R, Halldórsson S, et al. Ion mobility derived collision cross sections to support metabolomics applications. Anal Chem. 2014;86(8):3985–93.

6. Lanucara F, Holman SW, Gray CJ, Eyers CE. The power of ion mobility-mass spectrometry for structural characterization and the study of conformational dynamics. Nat Chem. 2014;6(4):281–94.

7. Zheng X, Aly NA, Zhou Y, Dupuis KT, Bilbao A, Paurus VL, et al. A structural examination and collision cross section database for over 500 metabolites and xenobiotics using drift tube ion mobility spectrometry. Chem Sci. 2017;8:7724–36.

8. May JC, Goodwin CR, Lareau NM, Leaptrot KL, Morris CB, Kurulugama RT, et al. Conformational ordering of biomolecules in the gas phase: Nitrogen collision cross sections measured on a prototype high resolution drift tube ion mobility-mass

13

spectrometer. Anal Chem. 2014;86(4):2107–16. 9. Paglia G, Astarita G. Metabolomics and lipidomics using traveling-wave ion mobility

mass spectrometry. Nat Protoc. 2017;12(4):797–813. 10. Paglia G, Angel P, Williams JP, Richardson K, Olivos HJ, Thompson JW, et al. Ion

mobility-derived collision cross section as an additional measure for lipid fingerprinting and identification. Anal Chem. 2015;87(2):1137–44.

11. Fenn LS, Kliman M, Mahsut A, Zhao SR, McLean JA. Characterizing ion mobility-mass spectrometry conformation space for the analysis of complex biological samples. Anal Bioanal Chem. 2009;394(1):235–44.

12. Stow SM, Causon TJ, Zheng X, Kurulugama RT, Mairinger T, May JC, et al. An Interlaboratory Evaluation of Drift Tube Ion Mobility-Mass Spectrometry Collision Cross Section Measurements. Anal Chem. 2017;89(17):9048–55.

13. Campuzano I, Bush MF, Robinson C V., Beaumont C, Richardson K, Kim H, et al. Structural characterization of drug-like compounds by ion mobility mass spectrometry: Comparison of theoretical and experimentally derived nitrogen collision cross sections. Anal Chem. 2012;84(2):1026–33.

14. Warnke S, Seo J, Boschmans J, Sobott F, Scrivens JH, Bleiholder C, et al. Protomers of benzocaine: Solvent and permittivity dependence. J Am Chem Soc. 2015;137(12):4236–42.

15. Metz TO, Baker ES, Schymanski EL, Renslow RS, Thomas DG, Causon TJ, et al. Integrating ion mobility spectrometry into mass spectrometry-based exposome measurements: what can it add and how far can it go? Bioanalysis. 2017;9(1):81–98.

16. Zhou Z, Shen X, Tu J, Zhu Z-J. Large-Scale Prediction of Collision Cross-Section Values for Metabolites in Ion Mobility-Mass Spectrometry. Anal Chem. 2016;88(22):11084–91.

17. Yuan M, Breitkopf SB, Yang X, Asara JM. A positive/negative ion-switching, targeted mass spectrometry-based metabolomics platform for bodily fluids, cells, and fresh and fixed tissue. Nat Protoc. 2012;7(5):872–81.

18. Pacholarz KJ, Barran PE. Distinguishing Loss of Structure from Subunit Dissociation for Protein Complexes with Variable Temperature Ion Mobility Mass Spectrometry. Anal Chem. 2015;87(12).

19. Surman AJ, Robbins PJ, Ujma J, Zheng Q, Barran PE, Cronin L. Sizing and Discovery of Nanosized Polyoxometalate Clusters by Mass Spectrometry. J Am Chem Soc. 2016;138(11).

20. Kurulugama RT, Darland E, Kuhlmann F, Stafford G, Fjeldsted J. Evaluation of drift gas selection in complex sample analyses using a high performance drift tube ion mobility-QTOF mass spectrometer. Analyst. 2015 Sep;140(20):6834–44.

21. Ujma J, Giles K, Morris M, Barran PE. New High Resolution Ion Mobility Mass Spectrometer Capable of Measurements of Collision Cross Sections from 150 to 520 K. 2016;

22. Ujma J, De Cecco M, Chepelin O, Levene H, Moffat C, Pike SJ, et al. Shapes of supramolecular cages by ion mobility mass spectrometry. Vol. 48, Chemical Communications. 2012. p. 4423.

23. Gidden J, Bushnell JE, Bowers MT, Barbara S. Gas-Phase Conformations and Folding Energetics of. 2001;(18):5610–1.

24. May JC, Morris CB, McLean JA. Ion mobility collision cross section compendium. Anal Chem. 2017;89(2):1032–44.

14

15