Metabolomics by NMR at CERM...Tenori L 1, Hu X , Pantaleo P, Alterini B, Castelli G, Olivotto I,...
Transcript of Metabolomics by NMR at CERM...Tenori L 1, Hu X , Pantaleo P, Alterini B, Castelli G, Olivotto I,...
Metabolomics by
NMR at CERM
Claudio Luchinat
CERM/CIRMMP University of Florence
Molecular biology, cellular
biology and biophysics labs
GENEXPRESS Genetic expression laboratory
Mass Spectrometry
X-Ray Crystallography
*equipped with cryoprobes
Conference Room
Library
Workstations
ss700wb
600 500*
400 700*
ss850wb 950*
1200**
900* 700*
He Liquifier
Biobank
Relaxometer
(0.01-40 MHz)
600
CERM/CIRMMP Magnetic Resonance Center
Main node of the Structural Biology Instruct-ERIC infrastructure
Biobank Lab (2016)
Fully or partially for metabolomics
Competence Center “Ivano Bertini”
(July 7, 2015)
ss800
Cryo EM**
** due in 2019
2 routes to
metabolomics
Celiac Disease Metabolomics
Clusterization of Celiac and Healthy subject serum spectra
Bertini, I.; Calabrò, A.; De Carli, V.; Luchinat, C.; Nepi, S.; Porfirio, B.; Renzi, D.; Saccenti, E.;
Tenori, L. The metabonomic signature of celiac disease, J. Proteome Res. 2009, 8(1), 170
Accuracy > 90%
Celiac Disease Metabolomics
Clusterization of Celiac and Healthy subject serum spectra
and corresponding Follow-up
Bertini, I.; Calabrò, A.; De Carli, V.; Luchinat, C.; Nepi, S.; Porfirio, B.; Renzi, D.; Saccenti, E.;
Tenori, L. The metabonomic signature of celiac disease, J. Proteome Res. 2009, 8(1), 170
Accuracy > 90%
Celiac – Healthy Subjects – Cross: predicted Potential Celiac
Bernini P, Bertini I, Calabrò A, la Marca G, Lami G, Luchinat C,
Renzi D, Tenori L. Are patients with potential celiac disease
really potential? The answer of metabonomics. J. Proteome
Res. 2010
There exists a metabolic fingerprint of celiac disease
These alterations are present also in potential celiac subjects: so they precede the intestinal
damage
Potential CD largely shares the metabonomic signature of overt CD. Most metabolites found to
be significantly different between control and CD subjects
were also altered in potential CD. Our results suggest early institution of GFD in patients
with potential CD
Celiac Disease Metabolomics
A metabolomic perspective on coeliac disease, A Calabrò, E Gralka, C Luchinat , E Saccenti, L
Tenori. Autoimmune Diseases, 2014
Colorectal Cancer Metabolomics
Cross-validated results on the Training Set:
Sensitivity : 79.9%
Specificity: 76.4%
Accuracy: 78.5%
Univariate Cox Regression Analysis for the Validation
Set:
HR: 3.30 95% CI: 2.02 to 5.37
P: 1.75 ∙ 10-6
PLS-CA model: long survival, in
blue; short survival, in yellow
Serum samples from 139 HS and 155 patients with
mCRC, included in a prospective phase II study of 3rd
line treatment with cetuximab and irinotecan
We can discriminate healthy controls from mCRC with
almost 100% accuracy.
More importantly, we can predict the overall survival of
the patients
Bertini I, Cacciatore S, Jensen BV, Schou JV, Johansen JS, Kruhøffer M, Luchinat C,
Nielsen DL, Turano P., Cancer Res. 2012 Jan 1;72(1):356-64.
Breast cancer metabolomics
Healthy vs
Met
Accuracy 73.44%
Healthy vs
Post-op
Accuracy 75.80%
Post vs
Met
Accuracy 74.96%
NOESY
Healthy vs
Met
Accuracy 72.67%
Healthy vs
Post-op
Accuracy 70.00%
Post-op vs
Met
Accuracy 70.00%
CPMG
Classification between
Pre-Op and Metastatic
subjects.
Accuracy ~80%
Other comparisons
Oakman C, Tenori L, Claudino
WM, Cappadona S, Nepi S,
Battaglia A, Bernini P,
Zafarana E, Saccenti E,
Fornier M, Morris PG,
Biganzoli L, Luchinat C,
Bertini I, Di Leo A.
Ann Oncol. 2011,22,1295-1301.
MSKCC Project
Random Forest was used to derive a score for relapse prediction.
ROC curve for CPMG spectra is significantly better than Adjuvant on line (AUC < 0.80).
Mol Oncol. 2014 Aug 10. pii: S1574-7891(14)00167-7
Breast cancer
Confusion matrix in the test set:
EBC no relapse EBC relapse
EBC no relapse 70.8 29.2
EBC relapse 23.8 76.2
Test set: 42 EBC women without relapse
192 EBC women with relapse
Training set: 85 EBC women without relapse
109 MBC women
Samples from a multicentric study in South
East Asia
Hart CD, Vignoli A, Tenori L, Uy GL, Van To T, Adebamowo C, Hossain SM, Biganzoli L,
Risi E, LoveRR, Luchinat C, Di Leo A., Clin Cancer Res. 2017 Mar 15;23(6):1422-1431
Accuracy: 73.5%
Sensitivity Specificity Accuracy
CMD vs CMS 45.52% 68.29% 61.19%
NYHA1 vs NYHA 2 61.88% 71.42% 67.71%
NYHA2 vs NYHA 3/4 73.62% 56.44% 68.04%
NYHA 1 vs NYHA 3/4 74.83% 68.55% 72.15%
Classification between different subgroups of Heart failure
patients (1D CPMG spectra).
Patients are separated from healthy, but there is not any significant
difference between the disease grading that could reflect the clinical
severity of the disease.
Although good discrimination between healthy and HF subjects with a severe
disease, if not expected, was easy to be hypothesized, a comparable good
discrimination ability between healthy and HF subjects with a mild disease was
unexpected and appears rather counter-intuitive.
Heart failure metabolomics
Patients vs Healthy 85.11% 91.04% 87.29%
Tenori L1, Hu X, Pantaleo P, Alterini B, Castelli
G, Olivotto I, Bertini I, Luchinat C, Gensini GF.
Int J Cardiol. 2013 9, 168, 113-5.
Heart failure metabolomics
The model for prediction of heart failure was developed for each of the 3 kinds of available
spectra: cpmg, noesy and diffusion edited
753 new healthy samples (blood donors) were tested against each model.
20 subjects were predicted as heart failure subjects in all three kinds of spectra.
We were able to recall 11 of them for an echocardiographical screening
6 out of 11 showed altered parameters (to be published)
Patient ID
(Gender)
LVEDD
(mm) IVSd
(mm)
LVEDD
Ind.
(mm/m2)
LVPWd
(mm)
LVM
ind
(gr/m2)
Aortic
root
(mm)
LA
(mm)
RV
(mm)
EF
(%)
2 routes to
metabolomics
No. of quantified metabolites
Sta
tist
ica
l
dis
trim
ina
tin
g p
ow
er
Fingerprinting vs. profiling
Fingerprinting
Profiling
So, why profiling?
• Portability (different NMR fields, NMR vs. MS, etc.)
• Biological insight
• Less prone to artifacts
• ...
B.I-LISA in Hepatitis
Diagnosis N°
HCV 67
HBV 50
HS 43
Defining the serum molecular fingerprint of
patients with liver disease of viral etiology
“Metabolic fingerprints between HCV and HBV patients: a possible interference of the two major hepatitis viruses in basal
metabolism pathways”
NMR
HCV HBV HS
HCV 83.1 14.1 2.8
HBV 10 58.6 31.4
HS 0 6.7 93.3
C-V 78.2 % predictive accuracy
OPLS-DA
OPLS-DA
OPLS-DA HCV HBV HS
HCV 69.3 15.4 15.3
HBV 13.9 58.8 27.3
HS 1.9 21.1 77
C-V 68% predictive accuracy
C-V 77% predictive accuracy
HCV HBV HS
HCV 81 10.8 8.2
HBV 15.8 64.4 19.8
HS 6.3 10.5 83.1
34 identified
metabolites in serum
samples
114 lipid sub-fractions
451 binning from NOESY
spectra
B.I-LISA in Hepatitis
EPIC study: breast cancer
The European Prospective Investigation into Cancer and Nutrition (EPIC)
study was designed to investigate the relationships between diet, nutritional
status, lifestyle and environmental factors, and the incidence of cancer and
other chronic diseases.
192 patients who developed breast cancer
96 High
mammographic
density
96 Low
mammographic
density
EPIC study: breast cancer
Low High
Low 61.5 38.5
High 36.3 63.7
Accuracy: 62.6%
P-value: <0.05
DIFFUSION spectra
Best discrimination using
diffusion spectra, due to the
contribution of lipoproteins.
BILISA found differences in TG,
especially in VLDL fractions.
EPIC study: breast cancer
(High Density/Low Density)
B.I.-LISA in Pathway-27
PILOT STUDY:
22 subjects with a diet enriched in DHA+BG
T0: before starting the diet
T1: after 4 weeks of diet
Serum samples analysed with B.I.-LISA
Pathway-27 is a pan-European interdisciplinary project addresses the role and
the mechanism of action of 3 bioactives (docosahexaenoic acid, β-glucan and
anthocyanins), chosen for their known effectiveness in reducing some risk
factors of metabolic syndrome, enriching 3 different food matrices (dairy-,
bakery-, and egg products).
B.I.-LISA in Pathway-27 p-value
T0
(median) T1
(median)
Tendency
in T1
Triglycerides 0.0198 193.56 153.63 ↓ mg/dL
Cholesterol 0.5144 252.825 250.905 mg/dL
LDL-Cholesterol 0.0456 136.435 147.095 ↑ mg/dL
HDL-Cholesterol 0.3288 50.415 51.505 mg/dL
Apo-A1 0.095 146.04 142.07 mg/dL
Apo-A2 0.0231 35.42 34.985 ↑ mg/dL
Apo-B100 0.9663 114.74 113.235 mg/dL
TRIGLYCERIDES
DISTRIBUTION p-value
T0
(median) T1
(median) Tendency
in T1
Triglycerides-VLDL 0.0198 130.675 112.03 ↓ mg/dL
Triglycerides-IDL 0.022 23.675 16.52 ↓ mg/dL
Triglycerides-LDL 0.3045 24.55 22.805 mg/dL
Triglycerides-HDL 0.0231 11.695 10.105 ↓ mg/dL
Triglycerides-VLDL-1 0.022 62.565 55.9 ↓ mg/dL
Triglycerides-VLDL-2 0.022 23.62 19.14 ↓ mg/dL
Triglycerides-VLDL-3 0.0231 18.71 16.97 ↓ mg/dL
Triglycerides-VLDL-4 0.0415 13.48 12.31 ↓ mg/dL
Triglycerides-VLDL-5 0.0469 3.59 3.315 ↓ mg/dL
Triglycerides-LDL-1 0.0446 6.845 6.1 ↓ mg/dL
Triglycerides-LDL-2 0.9693 2.3 2.145 mg/dL
Triglycerides-LDL-3 0.9663 2.985 2.955 mg/dL
Triglycerides-LDL-4 0.9271 3.005 3.035 mg/dL
Triglycerides-LDL-5 1 3.94 4.015 mg/dL
Triglycerides-LDL-6 0.5341 5.39 5.42 mg/dL
Triglycerides-HDL-1 0.3101 2.46 2.695 mg/dL
Triglycerides-HDL-2 0.0345 1.67 1.415 ↓ mg/dL
Triglycerides-HDL-3 0.0215 2.7 2.115 ↓ mg/dL
Triglycerides-HDL-4 0.022 4.615 4.015 ↓ mg/dL
METCOLON study
This project aims at developing a statistical model able to identify changes in
preoperative and postoperative metabolomic serum profiles of CC patients,
obtained via NMR, where loss of a cancer signal may be predictive of better
prognosis. We hypothesize that the metabolomic signal of cancer presence will
remain only if there is residual micrometastatic disease postoperatively.
METCOLON study
Accuracy: 80.0% Accuracy: 90.0%
Quantified Metabolites
(25 by Bruker IVDr) CPMG spectra (buckets)
METCOLON study
Higher at T1
T1/T0
METCOLON study
Accuracy: 80.0% Accuracy: 100.0%
T0
T1
T0
T1
Quantified Lipids (114) Diffusion spectra (buckets)
METCOLON study
LDL-Chol/HDL-Chol
Apo-B100/Apo-A1
Triglycerides LDL-6
Triglycerides LDL-5
VLDL Particle Number
Apo-B, VLDL
Free Cholesterol, HDL-2
Cholesterol, HDL-3
Apo-A1, HDL-3
Apo-A1, HDL-2
Phospholipids, HDL-3
Apo-A2, HDL-3
Higher at T1
T1/T0
Metabolomics at CERM Paola
Alessia
Cristina
Panteleimon Claudio
Veronica
Gaia
Leonardo
Doing research at CERM -
applications are welcome for:
Graduate student / Post-doc
Structural Biology/Metabolomics, ICT
projects, NMR developments
EMBO fellowships
Human Frontiers
Veronesi Foundation
Researcher (tenure or tenure track)
Accessing CERM infrastructure:
INSTRUCT
i-NEXT
EuroBioNMR
Phenomenal – e-infrastructure
WeNMR – GRID infrastructure