rapid diagnosis of parkinson’s disease from sebum using
TRANSCRIPT
doi.org/10.26434/chemrxiv.12517385.v1
Rapid Diagnosis of Parkinson’s Disease from Sebum using Paper SprayIonisation Ion Mobility Mass SpectrometryDepanjan Sarkar, Drupad Trivedi, Eleanor Sinclair, Sze Hway Lim, Caitlin Walton-Doyle, Kaneez Jafri, JoyMilne, Monty Silverdale, Perdita Barran
Submitted date: 19/06/2020 • Posted date: 26/06/2020Licence: CC BY-NC-ND 4.0Citation information: Sarkar, Depanjan; Trivedi, Drupad; Sinclair, Eleanor; Lim, Sze Hway; Walton-Doyle,Caitlin; Jafri, Kaneez; et al. (2020): Rapid Diagnosis of Parkinson’s Disease from Sebum using Paper SprayIonisation Ion Mobility Mass Spectrometry. ChemRxiv. Preprint.https://doi.org/10.26434/chemrxiv.12517385.v1
Parkinson’s disease (PD) is the second most common neurodegenerative disorder for which identification ofrobust biomarkers to complement clinical PD diagnosis would accelerate treatment options and help to stratifydisease progression. Here we demonstrate the use of paper spray ionisation coupled with ion mobility massspectrometry (PSI IM-MS) to determine diagnostic molecular features of PD in sebum. PSI IM-MS wasperformed directly from skin swabs, collected from 34 people with PD and 30 matched control subjects as atraining set and a further 91 samples from 5 different collection sites as a validation set. PSI IM-MS elucidates~ 4200 features from each individual and we report two classes of lipids (namely phosphatidylcholine andcardiolipin) that differ significantly in the sebum of people with PD. Putative metabolite annotations areobtained using tandem mass spectrometry experiments combined with accurate mass measurements.Sample preparation and PSI IM-MS analysis and diagnosis can be performed ~5 minutes per sample offeringa new route to for rapid and inexpensive confirmatory diagnosis of this disease.
File list (2)
download fileview on ChemRxivD.Sarkar PSI IM-MS Sebum Parkinsons disease.pdf (1.62 MiB)
download fileview on ChemRxivD.Sarkar PSI IM-MS Sebum Parkinsons disease SI.pdf (2.03 MiB)
1
Rapid Diagnosis of Parkinson’s Disease from Sebum using Paper
Spray Ionisation Ion Mobility Mass Spectrometry
Depanjan Sarkar1, Drupad K Trivedi1, Eleanor Sinclair1, Sze Hway Lim2, Caitlin Walton-Doyle1,
Kaneez Jafri1, Joy Milne1, Monty Silverdale2, and Perdita Barran1*
1Manchester Institute of Biotechnology, School of Chemistry, the University of Manchester,
Princess Street, Manchester, UK, M1 7DN.
2Department of Neurology, Salford Royal Foundation Trust, Manchester Academic Health
Science Centre, University of Manchester, UK.
Abstract
Parkinson’s disease (PD) is the second most common neurodegenerative disorder for
which identification of robust biomarkers to complement clinical PD diagnosis would
accelerate treatment options and help to stratify disease progression. Here we demonstrate
the use of paper spray ionisation coupled with ion mobility mass spectrometry (PSI IM-MS)
to determine diagnostic molecular features of PD in sebum. PSI IM-MS was performed
directly from skin swabs, collected from 34 people with PD and 30 matched control subjects
as a training set and a further 91 samples from 5 different collection sites as a validation set.
PSI IM-MS elucidates ~ 4200 features from each individual and we report two classes of
lipids (namely phosphatidylcholine and cardiolipin) that differ significantly in the sebum of
people with PD. Putative metabolite annotations are obtained using tandem mass
spectrometry experiments combined with accurate mass measurements. Sample
preparation and PSI IM-MS analysis and diagnosis can be performed ~5 minutes per sample
offering a new route to for rapid and inexpensive confirmatory diagnosis of this disease.
Introduction
Neurodegenerative diseases are the leading source of disability globally.1 According
to the 2015 Global Burden of Disease, Injuries, and Risk Factors Study (GBD), Parkinson’s
2
disease (PD) is globally the fastest growing neurological disorder.1 PD is also the second
most common age-related neurodegenerative disorder with a prevalence of approximately
2% among people aged 65 and over, with motor symptoms including bradykinesia, tremor,
rigidity, and postural instability. PD also has several non-motor symptoms, including:
hallucinations, dementia, depression, sleep disorders, constipation and loss of smell, among
others.2,3 The number of people suffering from PD worldwide has more than doubled
between 1990 and 2015 and is currently is estimated to be between six and ten million.1
Whilst some of this increase tracks the demographics of greater life expectancies, this does
not account for it entirely, and this increased occurrence is also due to better diagnosis and
to environmental factors, particularly in mid social security disability insurance (SSDI)
countries. Current predictions state that these numbers are projected to increase to over 20
million by 2050.1
Whilst the prevalence of PD increases with age, 4% of the total cases are under the
age of 50 and for this group the time from symptom development to diagnosis can be
longer, causing distress to the individuals and cost to healthcare systems.4 The increase in
PD globally, and its commensurate prevalence in younger people, compounds the need to
identify biomarkers and methods to detect them; to provide a diagnostic pathway that may
be applicable prior to the onset of motor symptoms.
One of the biggest unmet needs in the management of PD is the lack of disease
modifying drugs. One of the obstacles in drug development may be the identification of the
right PD population.5 Most clinical trials testing novel therapeutic agents recruit patients
who have been diagnosed with PD. We know that by the time PD patient develop motor
symptoms, extensive neuronal loss has already occurred.6 Neurodegeneration may be too
advance at this stage and the benefits of intervention are not seen even if the therapy is
effective. It is therefore crucial that there are reliable biomarkers to identify patients in the
pre-diagnostic/pre-motor phases of PD, before extensive neuronal loss occurs.
Developing oily or flaky skin, especially on the face and scalp, is a common symptom
of PD first noted in 1925.7-9 The light yellow, oily substance present on all human skin, is
known as sebum, and increased sebum production is a hallmark of PD. Sebum is produced
by the sebaceous glands in the skin, which help keep the skin and hair moisturized, and acts
3
to prevent sweat evaporating, thus assisting with the body’s temperature regulation. Sebum
is an underexplored biofluid, which is readily obtained from non-invasive skin swabs, that
primarily consists of a mixture of triglycerides, cholesterol, free fatty acids, waxy esters, and
squalene.10,11 We have previously shown that sebum contains volatile biomarkers of PD,12
and that it can reveal mitochondrial dysregulation as PD progresses.13 Here we set out to
develop a method to analyze sebum in its native state to facilitate rapid diagnosis of PD.
Paper spray ionization mass spectrometry (PSI MS) has already been used for the detection
of small molecules (50-800 Da) from unprocessed biofluids such as blood, urine, and CSF,14-
16 and represented a viable approach.
Combining ambient ionization with ion mobility mass spectrometry has merit as a
method to ‘clean up crude samples’ and provides reproducible ion selected drift time data
in the place of a retention time for identification along with m/z.17-20 Here we apply of paper
spray ionisation coupled with ion mobility mass spectrometry (PSI IM-MS) to the analysis of
sebum and demonstrate a-proof-of-principle application for PD biomarker discovery which
could be developed into a rapid clinical diagnostic test. PS IM-MS allows prompt sample
analysis with minimal sample processing/disruption, compared to other techniques such as
gas chromatography mass spectrometry (GC-MS) and liquid chromatography-mass
spectrometry (LC-MS), and separates analytes with overlapping m/z ratios which may be
isomers or isobarimers.
Results and Discussion
Optimisation of PSI MS for Sebum Analysis
The procedure for sample collection from sites across the UK (see Supplementary
information for further details on the sites) and subsequent PSI IM-MS analysis is shown in
Figure 1.
4
0 2000
Belfast
NORTHERNIRELAND
Scotland
Edinburgh
CardiffLondon
Wales
England
Mass spectrum collected from human sebum
m/z
kV
Recruitment clinics across the UK
Solvent
Sebum Collection
Transport
Transfer
Paper spray analysis
A B
C
D
E
F
Figure 1. Workflow for paper spray analysis from clinic to raw data A) Locations of the 28
collecting NHS clinics in the UK, B) sebum collection from the mid-back of participants, C)
medical Q-tips containing sebum samples transported under ambient conditions, D)
schematic of the sample transfer from Q-tip onto the paper, E) the paper spray process, and
F) representative mass spectrum collected from sebum with a distinctive multimodal
distribution.
Parameters of the PSI IM-MS experiments were optimised. These include: type, size and
shape of the paper, distance from the MS inlet, and eluting solvent composition. The size
and shape of the paper triangles was critical to achieving data with high precision, indicating
that the biomass of the sampling surface and the applied sebum can be controlled. Paper
triangles were cut manually, and each one was quality controlled for size prior to use. A
camera attached to the source housing ensured reproducible positioning of the paper tip for
each measurement. An optimized position was marked, and the paper tip placed at that
mark for each repeat. Whatman grade 1 and 42 filter papers were tested for PSI MS
experiments using standards (TM, L-proline, and L-glutamine). Following this optimisation,
5
the total ion chromatogram (TIC) and the mass spectra acquired using each filter paper were
visually similar (Figure S1), although reproducibility was higher with the Whatman grade 42
paper. Relative standard deviation (RSD) in the repeated measurements of the peak
intensity of L-glutamine was 9.6 using Whatman grade 42 paper in comparison to 15.06
using Whatman grade 1 paper (Figure S2); Whatman grade 42 paper was used for all further
analysis.
Crucial to reproducibility was sample transfer. Two methods were tested, firstly,
direct transfer to the paper triangle in a ‘touch and roll’ approach followed by recording PSI
MS from it and alternatively a rapid solvent extraction via vortex-mixing the sampled Q-tip
in ethanol (800 µL) for 5 s, followed by spotting on to paper for analysis. Figure S3 shows
mass spectra collected using these two approaches, both are rich in features, and notable is
the increased transmission of high molecular weight molecules (between m/z 1200-2000) in
the touch and roll transfer mass spectrum (Figure S3A) versus the solvent extraction
method, (Figure S3B). Hence the touch and roll approach was chosen for all further sebum
analyses using PSI MS. Different solvents and solvent mixtures were tested and H2O:EtOH
(v:v, 4:1) produced the best results and 4.5 μL of same was used to elute sebum for each
analysis. Following this optimisation, mass spectra of human sebum showed the presence of
three distinct envelopes of predominantly singly charged species in the higher mass region
(m/z 700-1800) (Figures 2A).
Improving Spectral Resolution with Drift Time Separation
Ion mobility-mass spectrometry (IM-MS) was used to further investigate these high
molecular weight metabolites, and specifically to resolve conformational isomers and
isobaric structural isomers as has been previously reported for lower molecular weight
lipids.21 Figure 2A and B shows the mass spectrum and a drift time (DT) vs. m/z plot of the
ions generated from a single PD patient sebum sample. For samples from both patients and
controls we find ~4150 m/z-DT features, of which 500 of these are found to be statistically
significant (p-value < 0.05). A subset of these statistically significant molecules have drift
time resolved features only present in PD samples and absent in controls. For example for
m/z 689.18 (Figure 2B grey box) and m/z 1394.83 (Figure 2B blue box) there is a
corresponding arrival time distribution (Figure S4). Separation between the two arrival
6
times for m/z 1394.83 was not to baseline on the Synapt G2-Si (Figure S4B) but this was
achieved using a Cyclic IM-MS (Waters Corp, UK) Figure S4C.
Figure 2C-F show three-dimensional DT vs. m/z plots from the m/z 700-900 region
for PD (blue boxes) and control (magenta boxes) samples, which contain regions that have
statistically significant differences, which here are evident even by visual inspection. A ridge
of doubly charged ions with arrival times of ~6.7 ms (indicated with a hollow red arrow in
Figure 2C and E) are common to the samples from PD and absent in controls (Figure 2C, E vs.
2D, F). Two later arriving ridges of ions at DT ~9.1 and ~10.5 ms show differences in relative
intensity between PD and control samples. The ions at ~6.7 ms are separated by 7.00 mass
units, and those at ~9.1 and 10.5 by 14.01 mass units as revealed across the m/z range by
drift time selected mass spectra, (Figure S5A, B) strongly suggesting that each contain
repeat units of CH2. These species that distinguish between PD and control samples are
observed across this envelope of ions (700-900 m/z) (Figure S6).
7
500 1000 1500 2000
Dri
ft ti
me
(ms)
m/z
C
E
D
F
DS_20190320_SEBUM DRUPAD_ION MOBILITY.raw : 1
m/z 689.18
AB
Se bum PD_01.raw : 1
Se bum PD_01.raw : 1
Se bum PD_01.raw : 1Se bum PD_01.raw : 1
m/z 1394.83
05
10
05
10
0
510
0
510
m/z
0
5
10
1550 1000 2000
PD samples Control samples
Figure 2. A) Mass spectrum collected using PSI IM-MS from a single person with PD. B) Drift
time vs. m/z plot with zoom regions exemplifying multiple DT peaks associated with a
specific m/z value at both low and high m/z that are deemed statistically significant. Three-
dimensional DT vs. m/z plots for PD (C and E) and control (D and F) samples from ~8 Da m/z
windows which exemplify the difference in the molecular composition of sebum produced
by people with PD. The arrows indicate the sets of drift time aligned features that alter in
intensity between PD and control. The open arrow shows the ridge of doubly charged
species with DT ~6.7, that are absent in controls and present in PD. The plots C-F are a
8
normalised average of samples from 81 (PD) and 74 (Control) from six recruiting sites (SI
Table 1).
Identification of Lipid Biomarkers of PD from Databases and with Tandem MS
In order to classify the statistically important distinguished features revealed in the
DT vs. m/z spectra, accurate mass searches in databases22,23 putatively identified multiple
classes of lipids, which matched predominantly to these as phosphatidylcholine and
cardiolipin families. Tandem mass spectrometry was used to increase confidence in these
putative annotations. Figure 3A-C show MS2 spectra for PC, PS, and CL class of lipid
standards, respectively. In each case, a diagnostic fragment ion was formed, which
corresponds to the mass of the polar head group for the respective lipid classes (Figure 3A-C
highlighted in red).
9
760.4 760.6 760.8 761.0
50 550 1050 1550
50 250 500 750 1000
50 250 500 750 1000
50 200 350 500 650 800 950
[M+H
]+
184.
08
760.61
184.08
810.57
208.00 [M+H
]+
20
8.0
0
1503.07
296.90
296.
90
[M+H
]+
865.77
202.23
202.
23
865
.77
839.75
202.23
83
9.7
5
202.
23
760.60
202.23
760.60
202.23
202.23
184.11
-H2O
20
2.2
3
76
0.6
0
MS2 of Lipid Standards MS2 of Lipids from Sebum
PC
PS
A
B
C
D
E
F
G
m/z 760.5990
m/z m/z
Inte
nsi
ty
CL
Figure 3. Tandem mass spectrometry data for standard lipids A) L-α-phosphatidylcholine
(PC), B) L-α-phosphatidylserine (sodium salt) (PS), and C) 18:1 cardiolipin (CL) showing
fragmentation of the head group as their fingerprint for identification. D-F) Show MSMS of
selected m/z values (760.00, 839.75, 865.77, respectively) from sebum samples. All these
selected ions fragment to m/z 202.23. G) MSMS spectrum for sebum in which the source
parameters gave in-source fragmentation to produce the m/z 202.23 fragment from its
precursor ions followed by isolation of the product ion for further fragmentation. Inset of D
shows a zoomed mass spectrum collected from sebum showing an accurate mass match
with a phosphatidylcholine with the chemical formula C42O8H83PN.
10
After obtaining fragmentation data from lipid standards, MS2 spectra were recorded for
sebum samples selecting different ions from the m/z 700-900 region. Figure 3D-F shows
three examples of species at m/z 760.00, 839.75, and 865.77 which were isolated and
subsequently fragmented using collision-induced dissociation (CID). In each case, a fragment
ion at m/z 202.23 was observed in the MS2 spectra, which could correspond to the aqueous
adduct of m/z 184.08 (the choline head group of PC). Hence, further investigation of m/z
202.23 was required to prove this speculation. An in-source fragmentation approach was
implemented to generate further fragments of the species at m/z 202.23. This species was
then mass isolated and fragmented using CID (Figure 3G). The presence of a peak at m/z
184.11 equates to a loss of 18.12 Da. from the head group of PC lipids. This data supports
the assignment of the fragment ion observed at m/z 202.23 as an aqueous adduct of the
choline head group of PC. Hence, the singly and doubly charged lipid molecules present in
the m/z 700-900 region following PSI MS of sebum are assigned to the phosphatidylcholine
lipid class. Accurate mass measurement data also supported this finding. Inset of Figure 3D
shows a zoomed view of a peak at m/z 760.5990 which corresponds to a
phosphatidylcholine molecule with chemical formula C42O8H83PN.
We also performed MS2 on higher molecular mass peaks. Figure S7 shows MS2
spectra for selected ions in the m/z 1500-1700 region (where here singly charged ions that
correspond to the doubly charged features shown in Figure 2C-F). The tandem mass spectra
yield fragment ions in the range m/z 750-900 region which is consistent with the
fragmentation pattern of standard CL (18:1 cardiolipin) (Figure 3C). The only difference
between the two is in the case of sebum, we see an array of fragment peaks in that region.
This observation can be attributed to the fact that sebum is a complex mixture of different
molecules. We hypothesise that it contains multiple CL with closely related chemical
structures that contribute to the array of fragment ions observed. Although the fragment
ion resembling the mass of the polar head group (m/z 296.90 in Figure 3C) was not visible in
the case of sebum, we contend that it may present as an adduct at a different m/z value. For
example, the fragment ion observed at m/z 365.29 (Figure S7D) can be assigned to [Head
group of CL+Na+K+3H2O]+. From the above data, we surmise that PC and CL lipid classes are
11
major components of sebum that can be elucidated using PSI-IM-MS, and that their
abundance is altered in the sebum of people with PD compared to matched controls.
Conclusion:
In conclusion, a rapid diagnostic test for PD from sebum samples has been
developed. PSI-IM-MS of each sebum sample is performed in ~ 2-3 min noticeably faster
than any other mass spectrometry based diagnostic methods such as, LC-MS or TD-GC-MS.
Whilst previous studies have demonstrated the use of PSI-MS for the detection of
metabolites present in blood, urine and other biofluids,14-16 it has not before been applied
to sebum. These results coupled with the lower signal from high mass species following
solvent extraction also demonstrate the benefit of such an ambient ionisation approach
direct from the native biofluid. LC-MS was not able to detect such large molecular weight
species (> 1200 Da) from sebum extracts.13
Mass spectra of sebum samples acquired using PSI-MS demonstrated the utility of
the technique for measuring both low and high molecular weight species (m/z 50-2000) that
may be lost in sample preparation, namely solvent extraction, in more traditional analytical
methods. PSI-MS combined with IM separation reveals specific compounds unique to PD
sebum samples when compared to healthy controls. Furthermore, we have identified two
classes of lipids namely phosphatidylcholine and cardiolipins, as components of human
sebum that are significantly differentially expressed in PD. Non-invasive sampling, rapid PS
IM-MS that targets these compounds could provide an inexpensive assay to support clinical
phenotyping for the confirmatory diagnosis of Parkinson’s Disease.
Acknowledgements
We thank The Michael J Fox Foundation (grant ref: 12921) and Parkinson’s UK (grant
ref: K-1504) for funding this study. D.S. thanks the Royal Society and SERB for the Newton
International fellowship. The Engineering and Physical Sciences and Biological and
Biotechnological Research Councils UK are also acknowledged for funding equipment used
in this study (EP/T019328/1 and BB/L015048/1). We are grateful to Tilo Kunath and Richard
Weller (University of Edinburgh) for helpful discussion on sebum and skin. We also thank
our recruitment centres for their enthusiasm and rigor during the recruitment process. We
12
are very grateful to all the participants who took part in this study as well as PIs and nurses
across all the recruiting centres.
Materials and Methods
For paper spray measurements, grade 1 and 42 filter papers were used (Whatman
International Ltd., UK). The paper was cut into isosceles triangles 5 mm (base) x 10 mm
(height). LC-MS grade solvents were used for the study. This included acetonitrile (98%
purity), methanol (99% purity) and deionized water (Fisher Scientific, UK) and ethanol (99%
purity) (VWR Chemicals, UK). Solvents were used without any further purification. An ESI-L
low concentration tuning mix (TM) (Agilent Technologies, UK), L-glutamine, and L-proline
(Sigma-Aldrich, UK) were used as standards for optimization of the process. For tandem
mass spectrometric experiments, a range of commercially available natural lipids were used
(Avanti Polar Lipids, Inc., USA): L-α-phosphatidylcholine (brain, porcine) (PC), L-α-
phosphatidylserine (brain, porcine) (sodium salt) (PS), and 1',3'-bis[1,2-dioleoyl-sn-glycero-
3-phospho]-glycerol (sodium salt) (18:1 cardiolipin) (CL). Tandem mass (MS2) spectra were
recorded for these lipids using PSI-MS. Solutions (1 mM) of PC in CHCI3:MeOH (50:50, v/v),
PS in CHCl3, and CL in MeOH were used for tandem mass spectrometric measurements. A
source was designed in-house (using Autodesk Inventor 2018) and 3D-printed (Ultimaker 3
Extended, GoPrint3D, Ripon, UK) for paper spray analysis on a Waters Synapt G2-Si HDMS
mass spectrometer. Copper micro-alligator clips (Premier Farnell UK Ltd., UK) were used to
hold the paper triangles at a high potential, followed by positioning it close to the MS inlet.
Medical Q-tips swabs (Fisher Scientific, UK and Copan Diagnostics, USA) were used for
sample collection.
Study Participants
For initial method development of paper spray ionization mass spectrometry (PSI-
MS) using sebum, samples from healthy controls were used. The developed method was
further tested using samples from participants with PD. The participants for this study were
part of a recruitment process taking place at 27 NHS clinics all over the UK (Table S1). A
subset comprising, 34 people with PD and 30 matched control subjects, was measured as a
training set and a further 91 (47 PD and 44 control) samples from 5 different collection sites
13
were used as a validation set. Ethical approval for this project (IRAS project ID 191917) was
obtained from the NHS Health Research Authority (REC reference: 15/SW/0354).
Sample Collection
Sebum samples were non-invasively swabbed from the mid-back of participants with
medical Q-tip swabs. Each swab, secured in its individual holder, was transported under
ambient conditions in sealed envelopes to the central facility at the University of
Manchester, where they were stored at -80 ⁰C until analysis.
Paper spray ionization mass spectrometry (PSI MS)
Sebum was transferred from the Q-tip swabs onto the paper substrates by gentle
touch and roll of the swab on to the sampling area. After sample transfer, the paper triangle
was clipped onto the copper alligator clip using tweezers avoiding contamination. Each
copper clip was cleaned by ultrasonication in acetone before use. For each sample, a new
clip and tweezer was employed to avoid cross-contamination. The clip was connected to a
custom paper spray ion source built in-house, adapted to the Synapt G2 Si HDMS ion
mobility mass spectrometer. PSI MS measurements were commenced by positioning the
paper tip in front of the MS inlet using a movable xyz nESI stage and subsequently applying a
voltage (2.5-3 kV) to the alligator clip by adapting the ESI capillary voltage supply. Upon
elution with a polar solvent at that elevated potential, a spray plume of tiny charged
droplets was observable at the tip of the paper simultaneously with the appearance of ion
signal.
All mass spectra were recorded over the range of m/z 50-2000. The critical
instrument parameters for each PSI-MS experiment were: capillary voltage at 2.5 kV, source
temperature at 80oC, sampling cone at 30 V and source offset of 40 V. No desolvation or
cone gas was used. Mass spectra were recorded for two minutes at a scan rate of 2 sec per
scan. A total of 60 scans were used for further data analysis.
Use of internal standards
To check the reproducibility of paper spray across different samples, TM was used.
For these experiments, 3.5μL of the TM solution was spotted on paper triangles and air-
14
dried. Dried paper triangles were used for PSI MS measurements of sebum samples
following an identical method described in the previous paragraph.
Data processing
After recording IM MS data from all the participant samples under identical
conditions, the raw data were deconvolved using Progenesis QI (Waters, Wilmslow, UK).
Peak picking initially identifies accurate mass m/z values and coincident ion mobility drift
times (DT). These correlated m/z-DT features are then aligned, and area normalization
carried out with reference to the best candidate sample, within the entire data set, chosen
by default set of parameters. Peak picking limits were set to automatic with default noise
levels, to balance signal to noise ratio according to the data quality. Signal acquired before
0.1 min of infusion and after 1.4 min of infusion were discarded during processing to only
retain reproducible signal. For annotation, accurate mass features were extracted and mass
matched with both the Human Metabolome Database (HMDB) and LipidMaps.22,23
Validation of PS-IM-MS data and extension to other sites
To determine the significance of this approach we repeated the analysis with ten PD
samples from five additional collection sites, to investigate the influence of location or the
person who collected sebum on the data. Principal component analysis (PCA) (Figure S8),
support vector machine (SVM), and random forest (RF) modelling classified samples by
collection sites. No obvious separation was possible using PCA, as expected for complex
data, since PCA is an unsupervised method of dimension reduction. To employ supervised
learning, the samples were repeatedly split into training set (75%) and a test set (25%) for
100 times. Each time a model was trained on training set and then tested using test set. The
prediction output from each test was output as a confusion matrix. Finally, a confusion
matrix representing the average of 100 tests was reported (Table S2 and S3 for SVM and RF
models, respectively). This analysis indicates that this data cannot be classified by collection
site, further, since samples from different sites and patients were acquired on different
days, we surmise that PSI IM-MS can be applied to detect differences in the molecular
composition of sebum that can diagnose PD without influence from the sampling
environment nor batching effects.
15
TOC
PSI IM MS
PD Control
Sebum
References:
1 Fereshtehnejad, S.-M. et al. Burden of neurodegenerative diseases in the Eastern Mediterranean Region, 1990–2016: findings from the Global Burden of Disease Study 2016. European Journal of Neurology 26, 1252-1265, doi:10.1111/ene.13972 (2019).
2 Hassan, Q., Li, S., Ferrag, C. & Kerman, K. Electrochemical biosensors for the detection and study of α-synuclein related to Parkinson's disease - A review. Anal. Chim. Acta 1089, 32-39, doi:10.1016/j.aca.2019.09.013 (2019).
3 Moustafa, A. A. et al. Motor symptoms in Parkinson's disease: A unified framework. Neurosci Biobehav Rev 68, 727-740 (2016).
4 https://parkinsonsdisease.net/basics/statistics/ 5 Stocchi, F. & Olanow, C. W. Obstacles to the Development of a Neuroprotective Therapy for
Parkinson's Disease. Movement Disorders 28, 3-7, doi:10.1002/mds.25337 (2013). 6 Marek, K. & Jennings, D. Can we image premotor Parkinson disease? Neurology 72, S21-S26,
doi:10.1212/WNL.0b013e318198df97 (2009). 7 Krestin, D. The Seborrhoeic Facies as a Manifestation of Post-Encephalitic Parkinsonism and
Allied Disorders. QJM: An International Journal of Medicine os-21, 177-186, doi:10.1093/qjmed/os-21.81.177 (1927).
8 Ravn, A.-H., Thyssen, J. P. & Egeberg, A. Skin disorders in Parkinson's disease: potential biomarkers and risk factors. Clin., Cosmet. Invest. Dermatol. 10, 87-92, doi:10.2147/ccid.s130319 (2017).
9 Arsic Arsenijevic, V. S. et al. A laboratory-based study on patients with Parkinson’s disease and seborrheic dermatitis: the presence and density of Malassezia yeasts, their different species and enzymes production. BMC Dermatology 14, 5, doi:10.1186/1471-5945-14-5 (2014).
10 Stewart, M. E. & Downing, D. T. Chemistry and function of mammalian sebaceous lipids. Adv. Lipid Res. 24, 263-301 (1991).
11 Lovaszi, M., Szegedi, A., Torocsik, D., Szegedi, A. & Zouboulis, C. C. Sebaceous-immunobiology is orchestrated by sebum lipids. Dermatoendocrinol 9, e1375636 (2017).
12 Trivedi, D. K. et al. Discovery of Volatile Biomarkers of Parkinson's Disease from Sebum. ACS Cent. Sci. 5, 599-606, doi:10.1021/acscentsci.8b00879 (2019).
13 Sinclair, E. et al. Sebum: a window into dysregulation of mitochondrial metabolism in Parkinson's disease. ChemRxiv, 1-20 (2020).
16
14 Damon, D. E. et al. Direct Biofluid Analysis Using Hydrophobic Paper Spray Mass Spectrometry. Anal. Chem. (Washington, DC, U. S.) 88, 1878-1884, doi:10.1021/acs.analchem.5b04278 (2016).
15 Espy, R. D. et al. Paper Spray and Extraction Spray Mass Spectrometry for the Direct and Simultaneous Quantification of Eight Drugs of Abuse in Whole Blood. Anal. Chem. (Washington, DC, U. S.) 86, 7712-7718, doi:10.1021/ac5016408 (2014).
16 Michely, J. A., Meyer, M. R. & Maurer, H. H. Paper Spray Ionization Coupled to High Resolution Tandem Mass Spectrometry for Comprehensive Urine Drug Testing in Comparison to Liquid Chromatography-Coupled Techniques after Urine Precipitation or Dried Urine Spot Workup. Anal. Chem. (Washington, DC, U. S.) 89, 11779-11786, doi:10.1021/acs.analchem.7b03398 (2017).
17 Yan, C. et al. Real-Time Screening of Biocatalysts in Live Bacterial Colonies. J. Am. Chem. Soc. 139, 1408-1411, doi:10.1021/jacs.6b12165 (2017).
18 Myung, S. et al. Coupling Desorption Electrospray Ionization with Ion Mobility/Mass Spectrometry for Analysis of Protein Structure: Evidence for Desorption of Folded and Denatured States. J. Phys. Chem. B 110, 5045-5051, doi:10.1021/jp052663e (2006).
19 Jackson, A. T. et al. Microstructural and conformational studies of polyether copolymers. Int. J. Mass Spectrom. 238, 287-297, doi:10.1016/j.ijms.2004.09.025 (2004).
20 Manicke, N. E. & Belford, M. Separation of Opiate Isomers Using Electrospray Ionization and Paper Spray Coupled to High-Field Asymmetric Waveform Ion Mobility Spectrometry. J. Am. Soc. Mass Spectrom. 26, 701-705, doi:10.1007/s13361-015-1096-z (2015).
21 Leaptrot, K. L., May, J. C., Dodds, J. N. & McLean, J. A. Ion mobility conformational lipid atlas for high confidence lipidomics. Nat. Commun. 10, 1-9, doi:10.1038/s41467-019-08897-5 (2019).
22 Wishart, D. S. et al. HMDB 4.0: the human metabolome database for 2018. Nucleic Acids Res 46, D608-D617, doi:10.1093/nar/gkx1089 (2018).
23 Fahy, E. et al. Update of the LIPID MAPS comprehensive classification system for lipids. J Lipid Res 50 Suppl, S9-S14, doi:10.1194/jlr.R800095-JLR200 (2009).
download fileview on ChemRxivD.Sarkar PSI IM-MS Sebum Parkinsons disease.pdf (1.62 MiB)
Rapid Diagnosis of Parkinson’s Disease from Sebum using Paper
Spray Ionisation Ion Mobility Mass Spectrometry
Depanjan Sarkar1, Drupad K Trivedi1, Eleanor Sinclair1, Sze Hway Lim2, Caitlin Walton-Doyle1,
Kaneez Jafri1, Joy Milne1, Monty Silverdale2, and Perdita Barran*1
1Manchester Institute of Biotechnology, School of Chemistry, the University of Manchester, Princess Street, Manchester, UK, M1 7DN.
2Department of Neurology, Salford Royal Foundation Trust, Manchester Academic Health Science Centre, University of Manchester, UK.
Centre ID
Centre Name PI
01 Royal Bournemouth General Hospital Mary Smolen
02 Southern Health Foundation Trust Matthew Young
03 South Tees Hospitals NHS Foundation Trust Sarah Morris
04 Salford Royal NHS Foundation Trust Monty Silverdale
05 Nottingham University Hospitals Gillain Sare
06 Western General, Edinburgh Gordon Duncan
07 Hampshire Hospitals Foundation Trust Deborah Dellafera
08 Cambridge University Hospital Rachel Ahmed
09 Sheffield Rosie Clegg
10 Bury Judith Brooke
11 Royal Cornwall Hospitals Ali James
12 Salisbury Alpha Anthony
13 London Cheryl Pavel
14 London
15 Luton & Dunstable Yvynne Croucher
16 Portsmouth Catherine Edwards
17 Bath Elizabeth Whelan
18 North Tyneside/Northumbria Steve Dodds
19 MRC Centre for Regenerative Medicine, Edinburgh
Tilo Kunath
20 Seb Derm, Edinburgh Richard Walker
21 Amsterdam, NL Anouk Rijs
22 JDR, Manchester Dani Mounfield
23 Plymouth Catherine Pitman/Sandra Morgan
24 Sunderland Anita Rutkauskaite
25 Devon Rob James
26 Gateshead Bryony Storey
27 Newcastle upon Tyne Alison Sutherland
28 Imperial College Ruby Colley
Table S1. Details of the collecting sites in the UK and the lead PI at each site.
0.0 0.5 1.0 1.5 2.0
0.0
4.0x107
8.0x107
0.0
4.0x107
8.0x107
500 1000 1500 2000
0.0
4.0x106
8.0x106
0.0
4.0x106
8.0x106
TIC Average mass spectrum
Whatman 42
Whatman 1
Inte
nsi
ty
Time (min) m/z
A C
B D
Figure S1. A-B) Total ion chromatogram of TM using PSI MS from Whatman 42 and Whatman 1, respectively. C-D) Corresponding average mass spectra. The total ion chromatogram (TIC) and the mass spectra acquired using each filter paper were visually similar, although reproducibility was higher with the Whatman grade 42 paper.
0.00E+00
1.00E-01
2.00E-01
3.00E-01
4.00E-01
5.00E-01
6.00E-01
1 2 3 4 5 6 7 8 9 10
Whatman 1
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18
1 2 3 4 5 6 7 8 9 10
Whatman 42
Figure S2. Reproducibility test using a set of 10 samples (L-glutamine under identical conditions), from Whatman 1 and 42.
0 500 1000 1500 2000
0 500 1000 1500 2000
PSI MS of sebum from
touch and roll transfer
PSI MS of sebum from
quick extraction in EtOH
m/z
Nor
mal
ized
inte
nsity
A
B
Figure S3. Mass spectra collected from sebum using A) touch and roll transfer and B) quick extraction
in 100% EtOH, indicating the presence of higher mass molecules (in between m/z 1200-2000) in case
of touch and roll transfer.
689.18
Time (a. u.)
11.0 11.5 12.0 12.5 13.0 13.5 14.0
1394.83
17-201901-2006 PD sample, HDMS, 1 pass
Time-0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00
%
0
100
-0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00
%
0
100
-0.00 2.00 4.00 6.00 8.00 10.00 12.00 14.00 16.00
%
0
100
SebumPD_01_dt_01 TOF MS ES+ 1394.803
3.85e3
12.32
9.69
DS_20190502_SAMPLE 12_dt_02 TOF MS ES+ TIC
3.74e7
3.25
1.80
1.25
4.36
SebumPD_01_dt_01 TOF MS ES+ TIC
3.84e6
2.97
1.53
1.19
5.69
4.42
A B
C1394.83
Figure S4. Arrival time distributions for single m/z values 689.18 (A) and 1394.83 (B). C)
Arrival time distribution for m/z 1394.18 using high resolution Cyclic IM-MS instrument.
800 850 900 950 1000 780 800 820 840 860 880 900
14.01
14.01
14.01
14.01
7.00
7.00
7.00
7.00
7.00
m/zm/z
BA
Figure S5. Drift time selected mass spectra from drift scope for the feature occurring at
observed at 14.01 Da. (A) and 7.00 Da. (B, being doubly charged) apart during PSI MS of
sebum.
05
100
510
0
510
0
510
PD samples Control samples
Figure S6. Three dimensional DT vs m/z plots for PD (blue boxes) and control (magenta
boxes) samples for other regions of statistical significance from m/z 700-900. The hollow
arrow indicates the region, drift time ~6.7 ms, where there are a series of doubly charged
ions that are only present with significant intensity in the PD samples. The two solid arrows
indicate the later arriving singly charged species that also present with significantly different
intensities between PD and control. The plots are a normalised average of samples from 81
(PD) and 74 (Control) from six recruiting sites (Table S1).
0 500 1000 1500 2000
0 500 1000 1500 2000
0 500 1000 1500 2000
15
03
.28
15
92
.38
16
18
.49
m/z
A
B
C
250 300 350 400 450 500 550
D
36
5.2
9
Inte
nsi
ty
Figure S7. A-C) MS2 spectra for selected ions in the m/z 1500-1700 region. D) Zoomed MSMS
spectrum following isolation of m/z 1592.38.
Site 4
Site 10
Site 11
Site 12
Site 16
Figure S8. Principal component analysis plot for PD samples from 5 different sites (10
samples from each site) to prove that the data was not influenced by different sites.
Predicted
Site 4
Actual Site
4
Actual Site
10
Actual Site
11
Actual Site
12
Actual Site
15
Predicted
Site 10
Predicted
Site 11
Predicted
Site 12
Predicted
Site 15
Table S2. Confusion matrix using support vector machine (SVM) model showing how many times
it was possible to predict a site correctly.
Predicted
Site 4
Predicted
Site 10
Predicted
Site 11
Predicted
Site 12
Predicted
Site 15
Actual Site
4
Actual Site
10
Actual Site
11
Actual Site
12
Actual Site
15
Table S3. Confusion matrix using random forest (RF) model showing how many times it was
possible to predict a site correctly.
download fileview on ChemRxivD.Sarkar PSI IM-MS Sebum Parkinsons disease SI.pdf (2.03 MiB)