Aiding diagnosis of rare disease: applications of mass spectrometry-based metabolomics in the Undiagnosed Diseases NetworkThomas O. Metz1, Jennifer E. Kyle1, Kelly G. Stratton2, Jean-Philippe Gourdine3, Erika M. Zink1, Young-Mo Kim1, Heino M. Heyman1, Lindsey N. Anderson1, Jeremy D. Zucker1, Bobbie-Jo M. Webb-Robertson2, Katrina M. Waters1, Melissa A. Haendel3, David M. Koeller41Biological Sciences Division, Pacific Northwest National Laboratory, Richland, WA USA ,2Computational & Statistical Analytics Division, Pacific Northwest National Laboratory, Richland, WA USA, 3Department of Medical Informatics and Clinical Epidemiology and OHSU Library, Oregon Health & Science University, Portland, OR USA, 4Department of Molecular & Medical Genetics, Oregon Health & Science University, Portland, OR USA

Introduction

Overview Methods Results

AcknowledgementsThis work was funded by NIH NCATS grant U01 TR001395. Samples were analyzed using capabilities developed in part by support of the NIH National Institute of General Medical Sciences (8 P41 GM103493-10) and the U.S. Department of Energy (DOE) Office of Biological and Environmental Research (OBER) Pan-omics program at PNNL. Metabolomics and lipidomics data were generated in the Environmental Molecular Sciences Laboratory, a DOE OBER national scientific user facility on the PNNL campus. PNNL is a multi-program national laboratory operated by Battelle for the DOE. The authors would like to thank Mr. Nathan Johnson for graphics support.

References1. Nakayasu et al., MPLEx: a Robust and Universal Protocol for Single-Sample

Integrative Proteomic, Metabolomic, and Lipidomic Analyses. mSystems. 2016

2. Burnum-Johnson et al., MPLEx: a method for simultaneous pathogen inactivation and extraction of samples for multi-omics profiling. Analyst. 2017

3. Kind et al., FiehnLib: mass spectral and retention index libraries for metabolomics based on quadrupole and time-of-flight gas chromatography/mass spectrometry. Anal Chem. 2009

4. Kyle et al., LIQUID: an-open source software for identifying lipids in LC-MS/MS-based lipidomics data. Bioinformatics. 2017

5. Dunn et al., Procedures for large-scale metabolic profiling of serum and plasma using gas chromatography and liquid chromatography coupled to mass spectrometry." Nat Prot. 2011

6. Imieliński et al., Investigating metabolite essentiality through genome-scale analysis of Escherichia coli production capabilities. Bioinformatics. 2005

Summary

Linking metabolomics and lipidomics results to patient genomes

CONTACT: Thomas O. Metz, Ph.D.Biological Sciences DivisionPacific Northwest National LaboratoryE-mail: thomas.metz@pnnl.gov

• 6% of the U.S. population suffers from a rare disorder that has evaded diagnosis

• We are performing mass spectrometry (MS)-based metabolomics analyses of plasma, cerebrospinal fluid (CSF), and urine from patients, as well as for Drosophila and zebrafish disease models

• To date, we have performed >2000 analyses of samples from healthy control individuals, patients and their first degree relatives

• >2000 metabolomics and lipidomics analyses performed, including of 391 individuals with no known metabolic disease and of 83 UDN patients and first degree relatives

• Reference data sets for plasma, urine, and CSF were generated

• Data from UDN patients and first degree relatives were normalized against and compared to the appropriate reference data sets in order to identify outlier metabolites or lipids

• A human metabolic knockout model was created to predict downstream metabolic effects of gene variants

• Work continues to integrate patient genome and metabolome/lipidome data• U.S. law defines a rare disease as one that affects < 200,000

individuals• To address undiagnosed disease, the NIH has established the

Undiagnosed Diseases Network (UDN; https://undiagnosed.hms.harvard.edu/)

• The UDN is comprised of a Coordinating Center, a Central Biorepository, 7 Clinical Sites, 2 Sequencing Cores, a Model Organisms Screening Center, and a Metabolomics Core (UDNMC) located at Pacific Northwest National Laboratory and Oregon Health & Science University (OHSU)

• The UDNMC is utilizing gas-chromatography (GC) and liquidchromatography (LC) coupled with mass spectrometry (MS) to perform untargeted metabolomics, lipidomics, and glycoproteomics analyses of samples from patients and their first degree relatives in the search for biomarkers

Reference sample and UDN participant demographics

omics.pnl.govCareer Opportunities: For potential openings in the Integrative Omics Group at PNNL please visit http://omics.pnl.gov/careers

Design• Patients accepted to the UDN have

exhausted all possibilities for a diagnosis• Once accepted, patients visit a clinical site

for a 1-week evaluation that includes collection of blood, urine, and if relevant, CSF samples

• To allow for proper statistical analysis of UDN patient data, we generated reference datasets from >391 individuals with no known metabolic disease and representative of UDN patient demographics

Approach• Metabolites, proteins, and lipids are simultaneously extracted from plasma

and CSF using the MPLEx protocol1,2. Metabolites and proteins are extracted from urine using methanol

• Metabolites are chemically derivatized, analyzed using GC-MS and identified based on match of retention indices and fragmentation spectra to an augmented version of FiehnLib3

• Lipids are analyzed using LC-MS/MS and identified based on their fragmentation spectra using the in-house tool LIQUID4

• Patient profiles are compared to reference datasets to identify outlier molecules using the workflow shown below5

Reference Datasets

Dr. Stanley Nelson of the UCLA Clinical Site performs various tests on a UDN patient during the 1-week evaluation. Photo source: Kaiser Health News

0%

10%

20%

30%

40%

50%

Plasma (136) Urine (102) CSF (149) UDN (601)

Num

ber o

f sub

ject

s

Sex

Female Male

Reference samples from individuals with no known metabolic disease were received from the Vanderbilt University Metabolic Screening Laboratory, the Mayo Clinic Biochemical Genetics Laboratory, the Oregon Clinical & Translational Research Institute Biolibrary, and the OHSU Layton Aging and Alzheimer’s Research Center.

Plasma metabolites

Plasma lipids

Analysis of samples and creation of reference data sets

Metabolite super classes Lipid subclasses

Comparison of UDN proband to reference data set

A human metabolic knockout model has been constructed to predict downstream effects

Overlaid chromatograms from 300 GC-MS-based metabolomics analyses (top frame) and 297 LC-MS-based lipidomics analyses (bottom frame) of reference and QC plasma samples. The chromatograms are shown prior to alignment.

Over 300 unique identified and unidentified metabolites and >500 unique identified lipids were measured in plasma, urine, and CSF samples from 391 reference individuals. Shown are the distributions of identified plasma metabolites and lipids among superclasses and subclasses, respectively.

Data from the reference samples are normalized and used to create reference data sets, against which data from UDN probands, and any relatives, are normalized and compared. Shown is a screenshot from a representative interactive z-score plot of plasma metabolomics data. The user can hover over individual points to display additional information.

Linking Gene Knockout with Metabolic Essentiality & Disease

Phenotype-Gene Relationships

PHOSPHOGLYCERATE KINASE 1 DEFICIENCY

PGK Pathway-Specific Metabolites

Phosphoglycerate kinase 1 deficiency

We have adapted a previously published metabolite essentiality model6 using VMH resource Recon2, in order to develop a genetically-based statistical model that includes all known reactions of intracellular small molecule biochemistry, including transport and core metabolism. The rate of change of each metabolite is determined by accumulation or depletion of a metabolite by other cellular processes intracellularly and extracellularly.

In this hypothetical example, the downstream metabolic effects of a deficiency in phosphoglycerate kinase 1 are predicted, as well as the likely associated patient phenotype.

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

0-0.5 yr 0.6-1.9 yr 2-10 yr 11-17 yr 18-30 yr 31-60 yr >60 yr

Perc

ent o

f sub

ject

s

Age

Plasma (136) Urine (102) CSF (149) UDN (601)

Aliphatic Acyclic Compounds (1)

Aliphatic Heteromonocyclic

Compounds (1)

Amino Acids, Peptides, and

Analogues (19)Aromatic

Heteromonocyclic Compounds (1)

Aromatic Heteropolycyclic Compounds (1)

Benzene and substituted

derivatives (1)

Benzenoids (1)

Carbohydrates and Carbohydrate

Conjugates (17)

Homogeneous Non-metal Compounds

(1)

Lipids (6)

Lipids and lipid-like molecules (1)

Nucleosides, Nucleotides, and

Analogues (1)

Organic acids and derivatives (18)

Organic Carbonic Acids and

Derivatives (2)

Organic nitrogen compounds (1)

Organoheterocyclic compounds (5) Carnitine, 4

Endocannabinoid (2) CE, 6

COQ10 (1)

Cer (15)HexCer (5)

LacCer (1)Gangliosides (5)

SM (35)

PC-lyso (23)

PC (88)

PC-O-Lyso (2)

PC-O (18)

PC-P (13)PE-lyso (7)

PE (26)

PE-O (6)

PE-P (40)PI-lyso (2)

PI (24)PS (1)

DG (6)

TG (137)

Metabolite Producibility MetabolitesTransport/Exchange MechanismsEnzymatic ReactionsPhenotypic Response

1) DATAVMH Human Metabolic Network Model composed of a curated collection of Genes, Metabolites, and Enzymatic Reactions associated with disease.

A

B

C

D

E

2DUptake RXN 1

RXN 4

F

Metabolic Function

3) ANALYSISIdentifying which genes are directly and/or indirectly responsible for each observed accumulation (+) and/or depletion (-) of a UDN metabolite as a result of the RXN KO.

2) MODELKnockout (KO) of all enzymatic reactions (RXN) one-by-one resulting in an accumulation (+) or depletion (-) of pathway-specific metabolites.

4) INTERPRETATIONMetabolite “C” is accumulated in the cytosol when “RXN 7” is knocked out; this reaction is associated with “Pubchem ID XXX/ OMIM: XXXX” involving Genes “ABA” and “ABC” of inherited disease.

Uptake R

XN 5