Rapid Metabolic Profiling of 1 μL Crude Cerebrospinal Fluid by Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry Imaging Can Differentiate De Novo Parkinson’s Disease

Parkinson’s disease (PD) is a highly prevalent neurodegenerative disorder affecting the motor system. However, the correct diagnosis of PD and atypical parkinsonism may be difficult with high clinical uncertainty. There is an urgent need to identify reliable biomarkers using high-throughput, molecular-specific methods to improve current diagnostics. Here, we present a matrix-assisted laser desorption/ionization mass spectrometry imaging method that requires minimal sample preparation and only 1 μL of crude cerebrospinal fluid (CSF). The method enables analysis of hundreds of samples in a single experiment while simultaneously detecting numerous metabolites with subppm mass accuracy. To test the method, we analyzed CSF samples from 12 de novo PD patients (that is, newly diagnosed and previously untreated) and 12 age-matched controls. Within the identified molecules, we found neurotransmitters and their metabolites such as γ-aminobutyric acid, 3-methoxytyramine, homovanillic acid, serotonin, histamine, amino acids, and metabolic intermediates. Limits of detection were estimated for multiple neurotransmitters with high linearity (R2 > 0.99) and sensitivity (as low as 16 pg/μL). Application of multivariate classification led to a highly significant (P < 0.001) model of PD prediction with a 100% classification rate, which was further thoroughly validated with a permutation test and univariate analysis. Molecules related to the neuromelanin pathway were found to be significantly increased in the PD group, indicated by their elevated relative intensities compared to the control group. Our method enables rapid detection of PD-related biomarkers in low sample volumes and could serve as a valuable tool in the development of robust PD diagnostics.


SUPPORTING FIGURES Page S2-S14
Overview of data Page S2 Figure S2 Structural validation of amino acids in CSF samples Page S3 Figure S3 Structural validation of neurotransmitters and other metabolites Page S4 Figure S4 MALDI-MSI of amino acids in CSF samples from PD patients and controls Page S5 Structural validation of neuromelanin-related metabolites Page S9-S10 Figure S9 Structural validation of m/z 163.087 with CE-MS/MS Page S11 Figure S10 Extracted ion electropherograms from CE-MS separations Page S12-S13 Figure S11 Effect of tobacco use on the investigated CSF metabolites Page S14

SUPPORTING TABLES Page S15-S20
Table S1 Limit of detection (LOD) of selected deuterated neurotransmitters and metabolites

Page S15
Table S2 Mass accuracy of isotope-labeled standards and corresponding endogenous NTs obtained from the LOD experiment Page S16 Table S3 List of synthetic standards used for structure validation of CSF detected metabolites and the theoretical m/z values of all potential derivatization species.
Page S17-S18   Representative mirror-matched MALDI tandem mass spectra of FMP-10 derivatized analytes in standard (black) and CSF (gray) samples collected with 20-30 V collision energy.Extracted peak lists from the corresponding MS/MS spectra can be found in Supplementary Data 2.

Figure S3. Structural validation of neurotransmitters and other metabolites.
Representative mirror-matched MALDI tandem mass spectra of FMP-10 derivatized analytes in standard (black) and CSF (gray) samples collected with 20-30 V collision energy.Extracted peak lists from the corresponding MS/MS spectra can be found in Supplementary Data 2. MALDI-MSI of selected FMP-10 derivatized neurotransmitters and metabolites in representative CSF spots from PD (blue), control (red) and quality control (grey) samples.All MSI data are normalized by root mean square (RMS) and scaled to 50% of the maximum intensity.
Step size 200 x 200 µm, laser size ~80 µm, 180 pixels/spot, runtime ~3 min/spot, mass resolution 220,000 at m/z 450.Information concerning molecular identification confidence and total number of identifications can be found in Table 1 and Supplementary Table 3. MALDI-MSI of selected FMP-10 derivatized neurotransmitters and metabolites on representative CSF spots from PD (blue), control (red) and quality control (grey) samples.All MSI data are normalized by root mean square (RMS) and scaled to 50% of the maximum intensity.
Figure S5MALDI-MSI of neurotransmitters and metabolites in CSF samples from PD patients and healthy controls Figure S1.Overview of data.(a) Box plots of age in the control and PD group.The mean age is indicated by + in each box plot.(b) Score plot of the two principal components of PCA including all variables after filtering and both technical replicas.The arrow indicates a strong outlier which was identified as an experimental error.(c) Score plot of the two principal components of PCA after the removal of the strong outlier.(d) Score plot of the two principal components of PCA after model optimization, i.e., removal of m/z values significantly different between the first and second technical duplicates.Abbreviations: PD, de novo Parkinson's disease patients; QC, quality control.

Figure S2 .
Figure S2.Structural validation of amino acids in CSF samples.

Figure S4 .
Figure S4.MALDI-MSI of amino acids in CSF samples from PD patients and controls.

Figure S5 .
Figure S5.MALDI-MSI of neurotransmitters and metabolites in CSF samples from PD patients and healthy controls.

Figure S6 .
Figure S6.Isomeric compound identification in CSF patient samples through MALDI-MS/MS analysis.Employing MALDI-MS/MS, isomeric compounds are delineated within CSF samples obtained from patients.This approach facilitated the precise characterization and differentiation of the L-dopa metabolite, 3-O-methyldopa (3-OMD), and its isobaric analog, the precursor of L-dopa, 3,4dihydroxyphenylalanine methyl ester (DHPME).The differentiation was based on thorough examination of MS/MS spectra in conjunction and the unique derivatization patterns of the two molecules.(a) Overlaid MALDI-MSI of m/z 479.197 (magenta, scaled to 20 % of maximum intensity) and m/z 746.301 (green, scaled to 20 % of maximum intensity) in CSF samples from control (circled in red) and PD (circled in blue) patients.(b) Structures of 3-OMD and DHPME.Derivatization sites are highlighted by gray circles.(c) Mirror-matched MALDI MS/MS spectra of FMP-10 single derivatized analytes (m/z 479.197) in standard (black) and CSF (gray) samples collected with 1 m/z isolation window and 30 V collision energy.(d) Mirror-matched MALDI tandem mass spectra of FMP-10 double derivatized analytes (m/z 746.301) in standard (black) and CSF (grey) samples collected with 1 m/z isolation window and 30 V collision energy.Step size 200 x 200 µm, laser size ~80 µm, 180 pixels/spot, runtime ~3 min/spot, mass resolution 220,000 at m/z 450.Extracted peak lists from the corresponding MS/MS spectra can be found in Supplementary Data 2. Abbreviations: 3-OMD, 3-O-methyldopa; DHPME, 3,4-dihydroxyphenylalanine methyl ester.

Figure S7 .
Figure S7.Validation of the PLS-DA model.(a) Permutation tests calculated from 100 permutations showing the correlation between the original and permuted statistics (R 2 , Q 2 ).The process included random re-ordering of the response variables (i.e., investigated classes), and the newly derived R 2 and Q 2 are plotted against the degree of correlation between the permuted and original data.The intercept is a measure of the overfit.Significantly lower R 2 and Q 2 values compared to the original values indicate the robustness of the model.Negative cumulative Q 2 (cum) values indicate that the ratio PRESS/SS, where PRESS is the squared differences between observed and predicted values for the Y-data kept out of the model fitting and SS is the residual sum of squares of the previous component, is higher than 1.(b) Correlation between the original scores of the PLS-DA model and the cross-validated scores for the first (PC1)and second (PC2) components.For both principal components, the correlation between the original and cross-validated scores was linear (R 2 >0.9), although better for PC1.

Figure S8 .
Figure S8.Structural validation of neuromelanin-related metabolites.(a) MS/MS imaging of precursor ion m/z 417.160, corresponding to the dehydrogenated single derivatized species of leukoaminochrome, also known as 5,6-dihydroxyindole (e), in CSF spots from one control (left) and one PD (right) sample.The product ion (m/z 389.165) exhibited a similar increase in the PD sample as the precursor ion.Step size 250 x 250 µm, laser size ~80 µm, 100 pixels/spot, runtime ~1.5 min/spot, mass resolution ~200,000 at m/z 417.(b) MS/MS imaging of precursor ion m/z 433.155, representing the single derivatized hydroxylated 5,6-dihydroxyindole in CSF spots from one control (left) and one PD (right) sample.The specific product ions (m/z 405.160, 403.144, 389.165) showed a similar increase in the PD sample as the precursor ion.All m/z images are scaled to 100% max intensity.Step size 200 x 200 µm, laser size ~80 µm, 180 pixels/spot, runtime ~3 min/spot, mass resolution ~200,000 at m/z 417.(c) Mirror-matched MS/MS spectra of precursor ion m/z 417.160 in FMP-10 derivatized leukoaminochrome standard (upper spectrum) and CSF (lower spectrum), collected with 1 m/z isolation window and 30 V collision energy.The CSF-specific product ion (m/z 389.164) is highlighted in both MS/MS spectra and is further exemplified in the enlarged, overlaid MS/MS spectra from the standard and CSF samples.The mirror-matched spectra display the % relative intensities of the detected ions within the mass range of m/z 200-420, applying a 1.5 signalto-noise ratio filter.The enlarged overlaid MS/MS spectra represent arbitrary intensity units.(d) Suggested fragmentation pattern of the singly derivatized species m/z 417.169 and m/z 433.155.The theoretical m/z values following derivatization are provided.(e) Oxidation of leukoaminochrome results in the CSF-detected species, which subsequently polymerize to form neuromelanin.

Figure S9 .
Figure S9.Structural validation of m/z 163.087 with CE-MS/MS.The MS/MS spectrum of norcotinine was obtained using a transfer collision energy ramp of 10-60 eV at 0.5 Hz and a precursor isolation width of ±0.015 Da.Prominent peaks at m/z 80.05 and 118.07 correspond to the published multiple reaction monitoring transitions commonly employed for the quantification of norcotinine using triple quadrupole MS/MS analysis.

Figure S10 .
Figure S10.Extracted ion electropherograms from CE-MS separations.(a)Extracted ion electropherograms (EIEs) obtained from CE-MS separations of two crude CSF samples.MALDI-MSI revealed the presence of the m/z 430.192 (ion image scaled to 50 % of max intensity) ion in sample 1, and subsequent analysis by CE-MS confirmed the presence of peaks corresponding to norcotinine, cotinine, and hydroxycotinine.The migration order and relative abundances observed under the experimental conditions were consistent with the reported levels of these metabolites in biofluids.Histidine and acetylcarnitine were utilized as reference compounds.When the m/z 430.192 ion was not detected in CSF using MALDI-MSI analysis, no putative nicotine metabolites were observed upon analysis by CE-MS.(b) EIEs reveal previously reported transitions of cotinine and hydroxynorcotinine resulting from in source decay of the parent ions during CE-MS analysis of crude CSF samples.Differences in migration time between panels (a) and (b) can be attributed to variances in separation potentials employed for the respective analyses.Step size 200 x 200 µm, laser size ~80 µm, 180 pixels/spot, runtime ~3 min/spot, mass resolution 220,000 at m/z 450.

Figure S11 .
Figure S11.Effect of tobacco use on the investigated CSF metabolites.(a) Distribution of tobacco users among control subjects and de novo Parkinson's disease (PD) patients based on age (y-axis) and gender (male/female).(b) Two-tailed t-tests (FDR < 0.05) were conducted to assess the impact of tobacco use on specific m/z values (x-axis) that exhibited significant differences between control and de novo PD samples.The y-axis represents the negative logarithm of the raw p-value, with magenta highlighting the m/z values that reached FDR < 0.05.

Table S2 .
Mass accuracy of isotope-labeled standards and corresponding endogenous NTs obtained from the LOD experiment.

Table S3 .
List of synthetic standards used for structure validation of CSF detected metabolites and the theoretical m/z values of all potential derivatization species.Only feasible derivatization species are included.

Table S4 .
Univariate statistical validation (two-tailed t-test) of the inspected high VIP m/z values.

Table S5 .
Untargeted univariate statistical analysis (two-tailed t-test) of the tobacco use effect.
* m/z values found significantly altered between de novo PD patients and control