Post on 23-Mar-2020
Intraoperative-DESI-MS-based Biopsy Analysis Platform for Glioma Diagnosis, Extent of Tumor Cell Infiltration Estimation,
and IDH SubtypingClint Alfaro, Valentina Pirro, Alan K. Jarmusch, Eyas M. Hattab, Aaron A. Cohen-Gadol,
R. Graham Cooks
Improving Health Through Researchindianactsi.org
MSACL 2019 Palm Springs, CA MS Imaging in Clinical Applications
1
Background and introductionOur goal: Develop and evaluate an intraoperativeDESI-MS-based tool to assist neurosurgery byproviding rapid diagnostic information
Gliomas
Claes et al. Acta Neuropathol ,2007, 114:443–458
-Surgical resection is primary treatment
-Grade IV five-year survival rate is 5%
-Residual tumor is primary cause of recurrence Intraoperative MS at IUSM
1) Isocitrate dehydrogenase (IDH) mutation status2) Disease status3) Tumor cell percentage (TCP)
2
Pirro et al. PNAS, 2017, 114(26): 6700-6705
Desorption electrospray ionization (DESI)
HV power supply
Nebulizer capillary
SurfaceSample
Desorbed ions
Freely moving sample stage in air
Atmospheric inlet of mass spectrometer
Ion transfer line
VSolvent
N2
Spraycapillary
Gas jet
Spray
Takats et al. Science 2004, 306, 471-473
Detects metabolites and lipids directly from fresh tissue
3
Workflow of the intraoperative DESI-MS experiment
4
1) Tissue Smearing 2) Sample Loading on DESI Source
3) Sample raster under DESI Spray 4) MS Data Acquisition
5
Tissue Smearing Sample Loading on DESI Source
Sample raster under DESI Spray MS Data Acquisition
Workflow of the intraoperative DESI-MS experiment
2HG Monitored with MS3 Product Ion Scan scan
2-hydroxyglutarate for assessing isocitrate dehydrogenase mutation status
IDH MutantIDH Wild-type
147
129
O
2HG DESI Signal Correlates with 2HG Concentration
α-ketoglutaric acid
2-hydroxyglutaric acid (2HG)
Mutant IDH
IDH Wild-type
y = 0.0619x - 2.4573R² = 0.986
0
10
20
30
40
0 250 5002H
G D
ESI-
MS
Sign
al
2HG Concentration (ng/mg tissue) from
Adjacent Tissue
IDH Mutant
Alfaro, C.M. et al. J. Neurosurgery 2019
Histopathology
Pathology-based
Region-Of-Interest
(ROI) Selection
Jarmusch et al. PNAS, 2016, 113(6): 1486-1491
Histopathologically Correlated DESI-MS Database was created
m/z 888m/z 885m/z 834
41 Tissue Sections
Grey matter
White matter
Glioma
m/z 281 m/z 303 m/z 788
m/z 834 m/z 885 m/z 888
Grey matter
Glioma
White matter
174
834
888
794
89885
Lipid and metabolite profiling with DESI-MS imaging
Data fusion
X
Y
Grey matter
Glioma
White matter
447 ROIs Selected
Assessing disease status using PCA-LDA
-1
7.5
16
80 140 200m/z
-1
7
15
700 850 1000m/z
Standard normal variate normalized metabolite
and lipid profiles
Met. PC Scores
Lip. PC Scores
Multiply by training set lipid and metabolite
loading matrices
Combined PC Scores
Met. profile
Lip. profile
Multiply by training set fused
loading matrix
Fusion PC ScoresLDA Prediction andClassification
Linear DiscriminantAnalysis
8
Grey matter
Glioma
White matter
Data fusion
Project unknowns onto fused PC score space and predict disease state with LDA
Intraoperative (unknown)
sample
SNV
No
rmal
ized
Inte
nsi
ty
Assessment of tumor infiltrationWhat about mixtures (real samples)?Do dynamics of lipid profile reflect tumor infiltration?
9
Glioma
White Matter
Grey Matter
Pirro et al. PNAS, 2017, 114(26): 6700-6705
Assess tumor infiltration using the lipid signature of the tissue
10
Y = bGMXGM + bWMXWM + bGXG
Regression using PC scores and ternary mixture model
[PC1 PC2] * [%GM %WM %GL]
Glioma
1:1:1 Mix
White Matter
Grey Matter
Pirro et al. PNAS, 2017, 114(26): 6700-6705
Unknown Spectrum
[%GM %WM %GL]
[%GM %WM %GL] = 100%
G
GMWM
Summary of intraoperative patient cohort
11
Subjects
58 recruited
6 Clinical Screenfails
3 DESI Outliers
Biopsies 247 Total
Smears
334 Total
62 excluded
272 smears included in
statistics
Gender24 F
25 M
Age (years) 46 ± 16
IDH Status
23 Mutant
25 Wild-Type
1 Unknown
Glioma
Grade
1 Grade I
17 Grade II
6 Grade III
25 Grade IV
794834
885.6
0
10
20
30
700 850 1000
89.0
174.0
0
50
100
80 140 200
85.1
101.1
0
0.1
0.2
0.3
0.4
80 95 110
Biopsy 155
Biopsy 154
888.7
0
105
210
700 850 1000
89.0
174.0
0
20
40
60
80 140 200
85.0
101.0
0
0.1
0.2
80 95 110m/z
Inte
nsi
ty
A
B C D
E F G
147
129
O
147
129
O
Lipids 2HGMetabolites
Biopsy 155: Glioma, High TCP, IDH Mutant
Biopsy 154: Infiltrative margin(white matter), Low TCP, IDH Mutant
Diagnosis of tumor cores and surgical margins-example from a single subject
12
IDH Mutation status was classified with 92.4% accuracy
158 smears, 29 subjects
13
Decision line = 6.18
Sensitivity 86.6%Specificity 97.5%
Comparison of histopathology diagnoses with DESI-MS diagnoses
Lipid Deconvolution TCP CategoryLow Medium High Total
Histopathology TCP Category
Low 92 23 11 126Medium 15 9 14 38
High 8 20 80 108Total 115 52 105 272
G = GliomaGM = Grey matterWM = White matterI.M. = Infiltrative margin
Low = 0-33% tumor cellsMedium = 34-66% tumor cellsHigh = 67-100% tumor cells
Mixed tissue samples are challenging for our current data and statistical methods14
PCA-LDADiagnosis from
DESI-MS
GI.M. (GM or WM)
Total
Histopathology DiagnosisG 94 43 137
I.M. (GM or WM) 20 115 135
Total 114 158 272
Conclusions
15
Intraoperative DESI-MS system was developed and utilized over 4-year period toanalyze brain tissue biopsies from 49 human subjects
IDH Mutation assay performed the best and provided accuracy of 92.4%
Accurate prediction of mixtures was difficult
Few low-TCP samples were misclassified as high-TCP; few high-TCP samples weremisclassified as low-TCP
Acknowledgements
NIH R21EB015722
16