Metabolic Signatures of Cystic Fibrosis Identified in Dried Blood Spots

Subscriber access provided by TULANE UNIVERSITY

Article

Metabolic Signatures of Cystic Fibrosis Identified in Dried Blood Spots For Newborn Screening Without Carrier Identification Alicia DiBattista, Nathan McIntosh, Monica Lamoureux, Osama Y Al-Dirbashi, Pranesh Chakraborty, and Philip Britz-McKibbin J. Proteome Res., Just Accepted Manuscript • DOI: 10.1021/acs.jproteome.8b00351 • Publication Date (Web): 03 Dec 2018 Downloaded from http://pubs.acs.org on December 5, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Journal of Proteome Research

Metabolic Signatures of Cystic Fibrosis Identified in Dried Blood Spots For Newborn Screening Without Carrier Identification

Alicia DiBattista1, Nathan McIntosh,2 Monica Lamoureux,2 Osama Y. Al-Dirbashi,2,3,4 Pranesh Chakraborty,2,3 Philip Britz-McKibbin1*

1

Department of Chemistry and Chemical Biology, McMaster University, Hamilton L8S 4M1,

Canada. 2

Department of Pediatrics, Children's Hospital of Eastern Ontario, Ottawa K1H 8L1, Canada.

3 Newborn

Screening Ontario, Children's Hospital of Eastern Ontario, Ottawa K1H 8L1,

Canada. 4

College of Medicine and Health Sciences, United Arab Emirates University, Al Ain 15551,

United Arab Emirates.

*Corresponding author: Philip Britz-McKibbin, E-mail: [email protected],Tel: +1-905-525-9140 x22771

Keywords: Metabolomics, cystic fibrosis, newborn screening, dried blood spots, CFTR, mass spectrometry, capillary electrophoresis, biomarker discovery, amino acids, ophthalmic acid

1 ACS Paragon Plus Environment

Journal of Proteome Research 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ABSTRACT Cystic fibrosis (CF) is a complex multi-organ disorder that is among the most common fatal genetic diseases benefiting from therapeutic interventions early in life. Newborn screening (NBS) for pre-symptomatic detection of CF currently relies on a two-stage immunoreactive trypsinogen (IRT) and cystic fibrosis transmembrane conductance regulator (CFTR) mutation panel algorithm that is sensitive, but not specific for identifying affected neonates with a low positive predictive value. For the first time, we report the discovery of a panel of CF-specific metabolites from a single 3.2 mm diameter dried blood spot (DBS) punch when using multisegment injection-capillary electrophoresis-mass spectrometry as a high throughput platform for nontargeted metabolite profiling from volume-restricted/bio-banked specimens with quality control. This retrospective case-control study design identified thirty-two metabolites, including a series of N-glycated amino acids, oxidized glutathione disulfide and nicotinamide that were differentially expressed in normal birth weight CF neonates without meconium ileus (n=36) as compared to gestational age/sex-matched screen-negative controls (n=44) after a false discovery rate adjustment (q < 0.05, FDR). Also, sixteen metabolites from DBS extracts allowed for discrimination of true CF cases from presumptive screen-positive carriers with one identified CFTR mutation and transient neonatal hypertrypsinogenemic neonates (n=72), who were later confirmed as unaffected due to a low sweat chloride (< 29 mM) test result. Importantly, six CFspecific biomarker candidates satisfying a Bonferroni adjustment (p < 7.25 E-5) from three independent batches of DBS specimens included several amino acids depleted in circulation (Tyr, Ser, Thr, Pro, Gly) likely reflecting protein maldigestion/malabsorption. Additionally, CF neonates had lower ophthalmic acid as an indicator of oxidative stress due to impaired glutathione efflux from exocrine/epithelial tissue, and elevation of an unknown trivalent peptide that was directly correlated with IRT (ρ = 0.332, p = 4.55 E-4). Structural elucidation of unknown metabolites was performed by high resolution MS/MS, whereas biomarker validation was realized when comparing a sub-set of metabolites from matching neonatal DBS specimens independently analyzed by direct infusion-MS/MS at an accredited NBS facility. This work sheds new light into the metabolic phenotype of CF early in life, which is required for better functional understanding of CFTR mutations of unknown clinical consequence and the development of more accurate yet cost effective strategies for CF screening.


Page 2 of 36

Page 3 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


INTRODUCTION Cystic fibrosis (CF) is one of the most common autosomal recessive disorders caused by mutations to the gene encoding the cystic fibrosis transmembrane conductance regulator (CFTR).1 More than 2,000 different mutations of CFTR comprising six broad mutation classes have been identified to date; however, most Caucasian CF patients are homozygous or heterozygous for an in-frame deletion of Phe at the 508 position (F508del).2 Although CF has long been considered a classic monogenic disorder, there is great clinical heterogeneity in phenotype as influenced by non-CFTR gene modifiers and environmental factors whose complex interactions with CFTR genotypes remain poorly understood.3 This contributes to variable symptom severity and disease progression reflecting the complex CF disease spectrum. Nevertheless, life expectancy and quality of life has significantly improved with early diagnosis since its recent inclusion within universal newborn screening (NBS) programs4, 5 as it allows for therapeutic interventions to be initiated before the onset of the first debilitating symptoms. Growing evidence has demonstrated the cost-effectiveness and efficacy of dietary interventions on later growth, lung function and survival for individuals diagnosed by NBS as compared to symptomatically for most infants without meconium ileus or CF family history.6, 7 Although a cure for CF still does not exist, early detection of CF by NBS together with standard CF clinical management (e.g., pancreatic enzymes, fat-soluble vitamins etc.) together with new targeted CFTR pharmacological therapies that restore chloride transport function8,

9

represent major

milestones to delay or prevent chronic pulmonary disease and lung transplantation later in life. A “two-tiered” strategy is most widely used for population-based CF screening based on detection of elevated immunoreactive trypsinogen (IRT) followed by a DNA mutation panel from a neonatal dried blood spot (DBS) collected around 2 days of age.10 Since elevated IRT levels are sensitive, but not specific for CF detection,11-12 a second-tier CFTR mutation panel is required to reduce false-positives while still using the original DBS specimen.13-15 However, the exact IRT threshold and total number of CFTR mutations included in panels varies widely by jurisdiction, which has contributed to diagnostic dilemmas in CF.16 Even when a floating cut-off is used for IRT to correct for antibody reagent lots and seasonal variations, the overall positive predictive value (PPV ≈ 4.5%) for CF screening by optimal IRT/DNA algorithms is low17 as compared to NBS for other genetic diseases by tandem mass spectrometry (MS/MS), such as 3 ACS Paragon Plus Environment


phenylketonuria.18 Limitations of CF screening include poor specificity due to transient neonatal hypertrypsinogenemia, the identification of unaffected carriers with CFTR allele variants who do not express the disease, and potential false-negatives for ethnic groups with mutations not included within panels.19 Alternative strategies have been proposed to improve PPV, including a repeat IRT20 or a pancreatitis-associated protein (PAP)21 plus DNA screen performed on a second DBS specimen collected from 7-10 days; however, these approaches contribute to higher health care costs while also delaying CF diagnosis. For these reasons, the pilocarpine-stimulated iontophoresis sweat test remains the “gold standard” for confirmatory diagnosis of all presumptive CF infants based on functional assessment of CFTR activity since sweat ducts normally display impermeability to chloride.22,

23

As a result, IRT and/or mutation screens

provide only probable information regarding disease risk with only about 5.6% of screenpositive CF infants having elevated sweat chloride (≥ 60 mM), whereas the majority of presumptive screen-positives (> 90%) are unaffected carriers or false-positives with low/normal chloride (< 29 mM).24 A direct consequence of CF screening, either with biochemical methods such as IRT or IRT/PAP with extended genetic sequencing is the identification of asymptomatic neonates with intermediate sweat chloride (30-59 mM) and an equivocal diagnosis.25 The frequency of these CF-screen positive inconclusive diagnosis (CF-SPID) cases are higher than anticipated based on evolving diagnostic and screening consensus guidelines.10 This raises the question - is there a better way to screen for CF that is ethical, cost-effective yet accurate to better inform clinical decision-making without undue parental anxiety? Since metabolites are “real-world” end-products of gene expression and environmental exposures, metabolomics offers a promising approach for biomarker discovery as required for new breakthroughs in precision medicine.26 Metabolomics not only serves to identify clinically relevant biomarkers of disease at early stages of latency,27 but also can provide new insights of the mechanisms of disease pathophysiology28 and reveal the phenotype of “silent” genetic mutations.29 As a result, there has been growing interest in applying metabolomics for characterizing the metabolic phenotype of CF.30 For instance, serum metabolomic studies using LC-MS and GC-MS revealed altered energetic metabolism in children with CF reflected by abnormal central energy metabolism, oxidative stress, and microflora activity.31 Similarly, metabolomic profiling of cultured human airway epithelial cells from CF and non-CF patients32 4 ACS Paragon Plus Environment

Page 4 of 36

Page 5 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


confirmed known perturbations in oxidative and osmotic stress. Volatile metabolites indicative of lung inflammation allowed for differentiation of CF children with and without positive cultures for P. aeruginosa based on analysis of their exhaled breath condensates.33 Also, metabolomic analysis of bronchoalveolar lavage (BAL) fluids demonstrated that branched-chain amino acids and lactate were associated with CF patients having airway inflammation.34 Similarly, metabolomic profiling on BAL fluids have shown that purines and amino acids were strongly correlated with neutrophil counts and lung function,35 which also served as predictive biomarkers of future lung disease.36 Recently, metabolomic studies of longitudinal sputum samples from adult CF patients revealed that platelet activating factor and related inflammatory lipids were elevated during active pulmonary exacerbations.37 However, metabolomic studies to date have thus largely focused on a mechanistic understanding of CF pathophysiology in patients with a confirmed diagnosis. Herein, we apply multi-segment injection-capillary electrophoresismass spectrometry (MSI-CE-MS) as a high throughput platform in metabolomics for biomarker discovery,38,

39

which revealed for the first time a panel of CF-specific metabolites measured

from DBS extracts of asymptomatic neonates without meconium ileus. METHODS Chemicals and Reagents. Ultra LC-MS grade methanol (Caledon, Georgetown, ON, Canada) and Ultra LC-MS grade acetonitrile (Honeywell, Muskegon, MI, USA) were used to prepare sheath liquid and BGE, respectively. Ammonium acetate, formic acid, 3-chloro-L-tyrosine (ClTyr), 3-chloro-L-tyrosine (Cl-Tyr), 2-[4-(2-hydroxyethyl)piperazin-1-yl]ethanesulfonic acid (HEPES), 4-fluoro-L-phenylalanine (F-Phe), 2-napthalenesulfonic acid (NMS) and all other chemical standards were purchased from Sigma-Aldrich (St. Louis, MO, USA). All stock solutions were made in distilled, deionized water and stored at 4°C, unless otherwise stated. CE-MS Instrumentation. Metabolomic studies on neonatal DBS extracts were performed on an Agilent 6550 QTOF-MS running Agilent MassHunter Workstation LC/MS Data Acquisition Software version B.05.01. All data processing was performed using MassHunter Qualitative Analysis Software version B.06.00. CE separations were performed with 30 kV of applied voltage at 25°C using uncoated fused-silica capillaries (Polymicro Technologies, AZ, USA) with 50 µM ID and a total length 110 cm. A background electrolyte (BGE) consisting of 1 M formic 5 ACS Paragon Plus Environment


acid, 15% v acetonitrile, pH 1.8 was used for the separation. The capillary was flushed with BGE at 950 mbar for 15 min between each run. Multiplexed separations were performed based on a serial hydrodynamic injection (50 mbar) alternating between a 5 s injection of sample (or QC) and a 40 s injection of BGE for a total of seven discrete samples analyzed within a single run.41 An Agilent 1260 Infinity isocratic pump was used to deliver a sheath liquid consisting of 60:40 MeOH:H2O with 0.1% formic acid supplied at a rate of 10 L/min. For real-time internal reference mass correction, the reference masses purine, hexamethoxyphosphazine and hexakis(2,2,3,3-tetraflurorpropoxy) phosphazine (HP-921) were spiked into the sheath liquid at 0.02% v to provide reference ions at m/z 121.0509 and m/z 922.0098. For analysis of all individual DBS samples, the QTOF-MS was operated as a full scan MS in 4GHz HiRes acquisition mode with a mass range of m/z 50-1700 and a scan rate of 1 Hz. The source parameters for the dual AJS ESI were as follows: dry gas = 16 L/min at 200°C, nebulizer = 8 psi, sheath gas = 3.5 L/min @ 199°C, VCap = 2000 V, Nozzle voltage = 2000 V, fragmentor = 380 V, skimmer = 65 V, Oct 1 RF = 750 V. At the start of each day, the QTOF was mass calibrated in positive ion mode and the spray chamber of the ion source and CE inlet electrode were cleaned with isopropanol:water (50:50) to prevent sample carryover and salt buildup as daily preventative maintenance followed by mass tuning of QTOF-MS instrument. Dried Blood Spot Extraction and Quality Control. Metabolites were extracted from a single 3.2 mm diameter circular DBS cut-out specimen (i.e., filter disks) by solvent extraction as optimized in our previous study.38 Briefly, disks were placed in 0.5 mL centrifuge tubes containing 100 µL of a 75% v methanol solution containing 10 M of F-Phe as an internal standard (IS). Disks were sonicated for 15 min and the extraction solution was then filtered through a Nanosep 3K Omega (3kDa MWCO) ultracentrifugation tube (Pall Life Sciences, MI, USA) at 14,000 g for 15 min to deproteinize samples. The filtrate was evaporated to dryness at room temperature in a Vacufuge vacuum concentrator (Eppendorf Inc., New York, USA). DBS extract filtrates were then reconstituted in 30 µL of distilled, deionized water containing 15% v acetonitrile, 10 µM Cl-Tyr as a secondary IS. A quality control (QC) and reference sample was prepared in-house by pooling together DBS extracts from one disk of each screen-negative neonates (n = 30) derived from the first batch of samples received. Separate QC aliquots were stored at -80 °C and each aliquot was thawed once prior to each analysis.


Page 6 of 36

Page 7 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Multiplexed Separations with Temporal Signal Pattern Recognition. A seven sample serial injection format in MSI-CE-MS was used as a high throughput metabolomics platform for biomarker discovery with high data fidelity.41 Various serial injection formats can be designed depending on the study design in order to encode mass spectral information temporally in the separation as previously outlined for characterization of the sweat metabolome from screenpositive CF infants.39 Briefly, duplicate injections (or pairs) of three DBS specimens prepared with a distinctive dilution pattern (1:2; 1:1, 2:1) were analyzed alongside a pooled QC from healthy neonates. Multiplexed separations were performed by MSI-CE-MS based on a serial hydrodynamic injection (50 mbar) program alternating between a 5 s injection of sample (or QC) and a 40 s injection of BGE for a total of seven discrete samples analyzed within a single run.39 All duplicate DBS extracts (including QC) were fully randomized in terms of their injection position within each run. This method was recently validated on proficiency DBS specimens from the CDC Newborn Screening Quality Assurance Program, as well as authentic neonatal DBS analyzed by stable-isotope dilution direct infusion-tandem mass spectrometry (DI-MS/MS) at Newborn Screening Ontario (NSO).38 Study Cohort and Metabolomics Data Workflow for DBS Extracts.

A retrospective

metabolomics study was performed by analysis of DBS extracts from a cohort of normal birth weight/gestational age CF neonates without meconium ileus (n=36) relative to screen-negative (SN)/healthy controls (n=44), and unaffected carriers or transient hypertrypsinogenemic cases (n=72). This study was approved by the CHEO Research Ethics Board and the Hamilton Integrated Research Ethics Board (REB#: 14-669-T) involving the secondary use of biobanked and de-identified specimens. CF diagnosis for affected infants was determined by a two-tiered IRT/DNA screening algorithm at NSO, and subsequently confirmed by sweat chloride testing (> 60 mM) at CF clinics at regional pediatric hospitals.42 In the province of Ontario, which has a CF incidence of about 1:3600, screen-positive (SP) infants with presumptive CF are currently classified within three categories according to both IRT levels and mutational status (Luminex xTAG CF39v2 Kit), namely category A (IRT > 96th percentile + 2 mutations) category B (IRT > 96th percentile + 1 mutation) and category C (IRT > 99.9th percentile + 0 mutations).43 Data from targeted analysis of metabolites previously identified in DBS extracts38 were combined with results from untargeted metabolite profiling using the MassHunter Molecular Feature Extractor 7 ACS Paragon Plus Environment


(MFE) algorithm.39 A minimum of 300 counts was used as a threshold for a molecular feature to be extracted to ensure that low abundance features were included and +H, +Na -H2O species for molecular ion were included. Manual filtering of known in-source fragments and adducts from a curated in-house list was also performed to exclude spurious and redundant signals. A molecular feature was then considered to be a unique sample-derived metabolite if it was not present in a blank DBS extract provided that it was reliably measured in QC samples with adequate precision (CV < 40%) throughout the study.39 All metabolites from DBS extracts were then defined by their characteristic mass-to-charge ratio and relative migration time (m/z:RMT) and included in the final data matrix. In this case, ion responses for each metabolite was normalized to F-Phe as an IS and reported as their integrated relative peak area (RPA). RPA for the duplicate injections were averaged to give a single measurement for each metabolite in every DBS sample. Nondetectable metabolites (i.e. < LOD) were replaced with half of the minimum RPA measured for that metabolite provided that they were detected in majority (> 75%) of individual neonatal DBS specimens in the study cohort. The original data derived from three batches of runs acquired over 18 months was then batch-corrected using the BatchCorrMetabolomics package44 in R-Studio (R-Studio) that takes advantage of a QC sample included in every run for MSI-CE-MS. This feature was critical when adjusting for long-term signal drift that is prone in ESI-MS even when implementing daily preventative maintenance and standard operating protocols.45 During completion of data acquisition for all three batches of DBS specimens in this study (May 2015, June 2016, October 2016), the QTOF-MS system underwent a service repair and detector replacement, which may have also contributed to instrumental drift over time. Metabolomics data from this study, including original data and batch-corrected data with QCs are included as an excel file in the Supporting Information. High Resolution MS/MS for Unknown Identification. High resolution, accurate tandem mass spectrometry (MS/MS) was used for structural identification of unknown metabolites based on reporting standards from the Metabolomics Standards Initiaitive.46 To ensure a strong precursor ion signal, CE-MS with on-line sample preconcentration via transient isotachophoresis/sample self-stacking was performed by injecting a long single plug of DBS extract for 90 s at 50 mbar.38 Prior to the injection, DBS extracts were diluted 2-fold in 400 mM ammonium acetate at pH 5.0. Precursor ions were fragmented by collision induced dissociation (CID) at three fixed collision energies of 10, 20 and 40 V. Either a CID averaged product ion spectrum or a product ion 8 ACS Paragon Plus Environment

Page 8 of 36

Page 9 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


spectrum at an optimal voltage was used for spectral interpretation. The Find by Targeted MS/MS algorithm in MassHunter was used to find putative compounds and subsequent structural elucidation was performed by searching the METLIN database via the MassHunter Personal Compound Database and Library (PCDL) manager with a minimum forward search score of 25 and a minimum reverse score of 80. If no database match existed in the METLIN PCDL, the compound was manually annotated using the fragment similarity and neutral loss search tools from the METLIN online database (http://metlin.scripps.edu), or compared with MS/MS spectra published in literature. Peaks Studio (Bioinformatic Studios Inc., Waterloo, ON, Canada) was also used as an in silico method for amino acid sequencing of an unknown peptide. MS/MS spectra for two unknown ions from neonatal DBS extracts subsequently identified in this study are included as “xml” files within a zip folder in the Supporting Information. Statistical Analysis. All electropherograms were processed using Igor Pro 5.0 software (Wavemetric Inc., Lake Oswego, OR, USA). Batch-corrected metabolomic data was logtransformed and autoscaled prior to multivariate statistical analysis, including unsupervised principal component analysis (PCA) and supervised partial least squares-discriminant analysis (PLS-DA) using Metaboanalyst 4.0.47 Non-parametric statistical methods, including MannWhitney U test, Kruskal-Wallis and Spearman rank correlation analysis were performed in SPSS/PASW 18 (SPSS Inc., PASW Statistics for Windows, Version 18, Chicago, IL, USA) as data was not normally distributed. Receiver operating characteristic (ROC) curves for top-ranked single or ratiometric metabolites were also performed on batch-corrected data using MedCalc (MedCalc Software, Version 12.5, Ostend, Belgium) as a binary classification model for discriminating true CF neonates from SN/healthy or presumptive CF yet unaffected neonates as evaluated by the area under the curve (AUC) and 95% confidence interval, as well as p-value.

RESULTS Study Design, Data Workflow and Batch Correction. Metabolomic studies were conducted using a validated MSI-CE-MS protocol and data workflow for biomarker discovery38 using a single circular 3.2 mm diameter circular DBS punch stored frozen (-80°C) at Newborn Screening Ontario (NSO) from 2012-2014. Table 1 summarizes the characteristics of a sex-balanced and normal birth weight cohort, where CF neonates were confirmed with two disease-causing CFTR 9 ACS Paragon Plus Environment


mutations (majority were homozygous or compound heterozygous for delF508; 34 of 36), high sweat chloride (> 60 mM) and pancreatic insufficient (35 of 36) based on fecal elastase-1 below cut-off levels (< 100 g/g). In contrast, other CF presumptive neonates who were unaffected (i.e. carriers and hypertrypsinogenemics, collectively referred to as SP/non-CF) had sweat chloride below cut-off levels (< 29 mM), whereas SN/healthy neonate controls had IRT concentrations within normal reference ranges. Most carriers were identified with one delF508 mutation (24 out of 37) or another less prevalent mutation, whereas CF infants were homozygous for delF508 (24 out of 36) or compound heterozygous with delF508 and another disease associated mutation (10 out of 36). No pre-term/critically ill infants were included in this study and all neonates were of normal gestational age (38-42 weeks) and birth weight (> 2500 g). Insert Table 1 Metabolomic studies were thus performed on three groups of neonates, namely confirmed CF cases, SN/healthy and SP/non-CF controls. All DBS extracts were randomly analyzed in duplicate by MSI-CE-MS, whereas a single QC was included in every run at different injection positions (Fig. 1A) when using full-scan MS acquisition under positive ion mode detection (Fig. 1B). A series of extracted ion electropherograms (EIE) for an unknown ion denoted by its unique m/z:RMT are depicted in Fig. 1C over a full day of analysis. Overall, 70 cationic metabolites were reliably detected (CV < 40% from QC runs) in the majority of DBS specimens (> 75% samples) after performing a dilution trend filter in MSI-CE-MS to reject degenerate signals, spurious peaks and background ions that contribute to data over-fitting and false discoveries in metabolomics.41 All known metabolites were identified by spiking with authentic standards (if available), whereas structural elucidation of unknown metabolites was performed by high resolution MS and MS/MS as shown for ophthalmic acid (OPA) in Fig. 1D and E. Insert Figure 1 A 2D scores plot from a principal component analysis (PCA) for the original dataset revealed three distinct clusters of QC (n=54) from each sample batch with poor technical precision as compared to the overall biological variance for all SP/presumptive CF (n=108) and 10 ACS Paragon Plus Environment

Page 10 of 36

Page 11 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


SN/healthy neonates (n=44) as shown in Fig. S1A of the Supporting Information. Furthermore, the extent of instrumental drift was metabolite-dependent as shown by representative QC control charts (n=54) for glycine (Gly), carnitine (C0) and N-glycated glutamic acid (Glc-Glu) plotted as a function of batch/date and injection number (Fig. S1B). However, batch-to-batch variability for both Gly and C0 was reduced significantly with a median CV < 10% following the QC-based batch correction algorithm while random variation for Glc-Glu was maintained (Fig. S1C). Overall, the experimental/batch-related variability in the repeat QC samples was reduced by 53%, whereas biological variation in the original data was largely preserved (Fig. S1D). Metabolic Phenotype of True CF Neonates as Compared to Screen-negatives. A PCA 2D scores plot is depicted in Fig. 2A, which provides an overview of the improved batch-corrected metabolomics data as reflected by a single tight cluster of QC samples centered within SN/healthy neonate group with a median CV of 8% indicative of excellent technical precision. As expected, individual DBS extracts from a cohort of SP/presumptive CF cases (n=108) and SN/healthy neonate controls (n=44) have much greater biological variance with a median CV of 34% based on 70 polar/ionic metabolites consistently detected in the majority of samples. The first stage of this study was aimed at identifying metabolites capable of differentiating between asymptomatic CF neonates and healthy neonates based on their characteristic metabolic phenotype from DBS specimens collected around 2 days after birth. Fig. S2 of the Supporting Information depicts the distinctive metabolic phenotype of CF neonates as compared to healthy/SN controls when using supervised multivariate statistical analysis based on partial leastsquares discriminate analysis (PLS-DA). However, since the relative ion response for metabolites were not always normally distributed following a log-transformation (Shapiro-Wilk p < 0.05), non-parametric statistical analysis was performed on non-transformed data. Overall, 32 metabolites were differentially expressed in CF neonates as compared to healthy controls using a Benjamini-Hochberg false discovery rate (FDR) correction (q < 0.05) as listed in Table 2. A series of N-glycated amino acids were tentatively identified by high resolution MS/MS based on their characteristic neutral losses (e.g., -C6O5H10, -H2O) as shown in Fig. S3 of the Supporting Information. Fig. 2B depicts box-whisker plots for the top-ranked metabolites associated with CF neonates, which included higher N-glycated glutamic acid (Glc-Glu), a glycated oxidized glutathione (Glc-GSSG), and nicotinamide, as well as lower glutamine (Gln),



tyrosine (Tyr), and O-acetyl-L-carnitine (C2) in DBS extracts. Additionally, Fig. 2C depicts receiver operating characteristic (ROC) curves for top-ranked single (Glc-Glu) and ratiometric (Thr/Glc-Glu) metabolites with excellent discrimination of CF from healthy neonates (p < 0.0001) with an area under the curve (AUC) > 0.920. Insert Figure 2 Insert Table 2 Fig. 3A depicts a full-scan TOF-MS spectra consistent for Glc-GSSG highlighting that the protonated molecule was a divalent ion [MH22+] with two sulfurs based on its isotopic pattern. Also, the addition of dithiothreitol (DTT) as a reducing agent to a DBS extract led to complete attenuation of the signal for the protonated molecule, confirming the unknown ion was a mixed oxidized glutathione disulfide with the formation of reduced glutathione (GSH, m/z 308.091, MH+) along with a new protonated molecule not detected in the original sample (m/z 470.144; MH+) (Fig. 3B). Furthermore, fragmentation of the precursor ion using MS/MS confirmed a neutral loss of a hexose (m/z 162.053) and a low signal for GSSG (m/z 307.083; MH22+) as a product ion. Collectively, accurate MS, MS/MS spectral assignments and thiolspecific chemical reactivity demonstrate that the unknown divalent ion was an N-glycated glycine mixed oxidized glutathione (Glc-GSSG) as shown in Fig. 3C. Overall, 17 of the 32 topranked metabolites (q < 0.05) associated with CF neonates also satisfied a Bonferroni adjustment (p < 7.25 E-5) with moderate (> 0.40) to strong (> 0.60) effect sizes as summarized in Table 2. Insert Figure 3 Differentiation of CF Neonates from Unaffected Screen-Positives/Carriers. We next explored the feasibility to differentiate true CF from SP/non-CF newborns due to the poor positive predictive value of conventional two-tiered IRT/DNA screening algorithms. Indeed, all unaffected carriers and transient hypertrypsinogenemic cases (without an identified mutation within CFTR panel) are designated as presumptive CF prior to confirmatory sweat chloride testing. Fig. S4 in the Supporting Information depicts a 2D scores plot from PLS-DA together with VIP score ranking of metabolites (VIP scores > 1.5) associated with CF neonates, which 12 ACS Paragon Plus Environment

Page 12 of 36

Page 13 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


were also confirmed by their box-whisker plots and ROC curves. Similarly, Table 3 summarizes the results from a Mann-Whitney U test (without data transformation), where 16 metabolites from DBS extracts were differentially expressed (q < 0.05, FDR) in CF neonates (n=36) as compared to unaffected SP/non-CF (n=72) controls, including 5 metabolites that satisfied a Bonferroni adjustment (p < 7.25 E-5) with moderate effect sizes (> 0.400). Insert Table 3 These CF-specific metabolites corresponded primarily to amino acids that were present at lower concentrations in CF neonates (e.g., Thr, Ser, Pro, Tyr) similar to results measured in healthy neonates (Table 2). Noteworthy, one of the most significant CF-specific biomarkers was OPA, a glutathione analog where cysteine is substituted with 2-aminobutyric acid, which was also lower in CF infants (fold-change, FC = 0.81; p = 6.52 E-5, effect size = 0.471). OPA was identified based on high resolution MS and lack of signal attenuation upon DTT addition (not a disulfide), whereas its MS/MS product ion spectrum (Fig. 1E) was consistent with data reported by Soga et al.48 Unambiguous OPA identification was also confirmed by direct comparison of MS/MS spectra acquired from a pooled neonatal DBS extract relative to an authentic commercial standard. Fig. S5 of the Supporting Information depicts a mirror plot for MS/MS spectra of OPA highlighting an excellent match based on a low mass error (1.7 ppm) for its protonated molecular ion [MH+], as well as three diagnostic fragment ions having similar relative intensity. Only one metabolite was significantly elevated in CF neonates (FC = 1.17; p = 3.58 E-3) and this corresponded to an unknown ion (m/z 438.258). Fig. S5 of the Supporting Information confirms that this CF-specific compound was a trivalent protonated molecule [MH33+] that was not reactive to DTT, and its MS/MS spectra was consistent with a polypeptide having a neutral mass of 1311.75 Da (C44H83N9O19). However, its exact molecular structure remains unclear following an extensive database search in open-access metabolomic spectral libraries (e.g., Metlin, HMDB). Further MS/MS spectral annotation together with in-silico prediction software confirmed that this peptide contained branched-chain amino acids with a putative amino acid sequence of VLSPCFLLRHQ (average local confidence = 67%) that resulted in no known human protein matches following a BLASTp search.



Table S1 summarizes the results of Kruskal-Wallis and pair-wise Mann-Whitney U tests when comparing true CF neonates with both SP/non-CF and SN/healthy controls, including 33 significant metabolites (q < 0.05) detected in DBS extracts by MSI-CE-MS, as well as IRT (ng/mL) measured independently by standard fluorescence-based immunoassays at NSO. Interestingly, all N-glycated amino acids (e.g., Glc-Glu, Glc-Gln, Glc-Leu/Ile, Glc-GSSG, GlcGly) and an unknown metabolite (m/z 162.076; MH+) were significantly elevated in both CF and SP/non-CF cases relative to healthy neonates, whereas certain amino acids (Orn, His, Gln), nicotinamide, GSSG and acetylcarnitine were depleted in all presumptive CF cases relative to SN/healthy neonates. All metabolites are annotated by their characteristic m/z:RMT as well as a confirmed or tentative compound ID, together with a mass error (ppm). Exact two-tailed pvalues and FDR corrected p-values are listed for the Kruskal-Wallis test, where metabolites satisfying a FDR correction (q < 0.05) in the Mann-Whitney U test are denoted in bold. Similar to Table 3, 17 metabolites and IRT (FC = 1.60; p = 7.00 E-7) were found to be differentially expressed in true CF relative to SP/non-CF neonates. Fig. 4A-C depicts box-whisker plots for OPA, Tyr and Ser (similar trends for other CF-specific amino acids, such as Thr, Pro, Ala, Gly and Gln), including a ROC curve based on OPA/unknown peptide ion ratio (Fig. 4D) that shows good differentiation of CF neonates from unaffected SP/non-CF cases with an AUC = 0.784 (p < 0.001). Intriguingly, the unknown peptide was the only metabolite measured in DBS extracts that was positively correlated with IRT based on a Spearman rank correlation analysis (ρ = 0.332, p = 4.55 E-4) as shown in Fig. 4E. Insert Figure 4 Targeted Panel of Biomarkers Already Screened by MS/MS. Several amino acids and acylcarnitines measured from DBS extracts by MSI-CE-MS in fact overlap with a targeted panel of 44 biomarkers (Table S2 of Supplemental Information) already implemented in NBS for multiplexed screening of twenty-two different inborn errors of metabolism. As a result, we retrospectively analyzed original data acquired by stable-isotope dilution DI-MS/MS49 within an accredited laboratory at NSO derived from matching DBS specimens processed within days of original collection as compared to stored DBS cut-outs (> 2 yrs at -80°C) from the same neonates when using MSI-CE-MS. Table S3 revealed that ten (out of 44) metabolites were 14 ACS Paragon Plus Environment

Page 14 of 36

Page 15 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


consistently lower in CF relative to healthy neonates (q < 0.05) as measured by DI-MS/MS, including three metabolites independently measured by MSI-CE-MS, namely Tyr, Gly and Arg (Table S1). The other medium-chain acylcarnitines (C12, C3DC, C4DC, C4OH and C18-2) were not detected by MSI-CE-MS as they were below the method detection limit when using full-scan data acquisition, whereas total leucines (Leu + Ile) and Phe were not found to be significant. Importantly, Table S4 demonstrates that only 6 (out of 44) metabolites from original DI-MS/MS data were significantly lower (q < 0.05) in DBS extracts from CF as compared to SP/non-CF neonates, including Gly, Tyr, Phe, Met, C8 and C2. Noteworthy, two amino acids (Gly, Tyr) also satisfied a Bonferroni adjustment (p < 0.00114) when correcting for multiple hypothesis testing with moderate effect sizes (0.500), including an average FC ≈ 0.80 consistent with outcomes measured by MSI-CE-MS on stored DBS specimens. Fig. S6 of the Supporting Information depicts an inter-laboratory method comparison for measured Tyr concentrations from matching DBS specimens analyzed from the full cohort (n=154) based on Passing-Bablok regression and Bland-Altman % difference plots. Overall, a mean bias of +6.9% for Tyr quantification was demonstrated between the two MS platforms with good mutual agreement as reflected by a slope (1.17) that did not deviate significantly from the line of unity (p > 0.05) with few outliers among a randomized data distribution (i.e., homoscedactic). Similar outcomes were achieved for Gly from matching DBS specimens for the first sample batch with a mean bias of -1.2% and slope of 0.911 (n=54) between the two different MS platforms (data not shown). As a result, there was no evidence of systematic error due to methodology, extraction procedure and/or long-term DBS storage. Importantly, a large fraction of metabolites measured by MSICE-MS are not included within current targeted biomarker panels by DI-MS/MS for NBS, such as OPA, Ser and Thr. DISCUSSION Despite conclusive data supporting the economic, social and long-term health benefits of universal NBS as compared to non-screened populations,50, 51 neonatal screening strategies for CF have presented unique ethical and technical challenges to public health programs. While there is no significant long-term psychosocial harm from false-positive results, many parents report transient anxiety/distress provided that delays are minimized from initial notification and confirmatory testing to reach a conclusive CF diagnosis.52 In this context, NBS for CF largely 15 ACS Paragon Plus Environment


adopts a two-tiered IRT/DNA screening algorithm followed by a confirmatory sweat test for all presumptive SP cases. Recently, the state of California evaluated a 4 yr pilot study incorporating next generation sequencing of the CFTR gene into the screening algorithm as a third-tier following IRT/DNA in order to improve detection of CF within a multiethnic population.53 However, as less than 20% of all CFTR variants have known or probable clinical consequence, extended genetic sequencing results in the detection of a large fraction of unaffected carriers with low sweat chloride, as well as CF-SPID cases with inconclusive diagnosis.25 However, the main goal of NBS is early detection and prompt therapeutic intervention of CF neonates, not identification of unaffected carriers for family planning, nor those at risk for late-onset/mild phenotypes who do not require treatment; thus current NBS algorithms for CF remain suboptimal for this purpose. In this work, we hypothesized that MS-based metabolic phenotyping may enable more accurate screening of affected CF neonates in the population. To the best of our knowledge, this is first comprehensive metabolomics study as applied towards differential screening of presumptive CF neonates using bio-banked DBS specimens. In our work, 29 metabolites (q < 0.05, FDR) measured from DBS extracts were differentially expressed in CF neonates relative to SN/healthy controls (Table 2), which included significant increases among a series of N-glycated amino acid adducts, as well as lower levels of GSSG, nicotinamide and several amino acids. The N-glycated amino acids likely represent markers of hyperglycemia formed by non-enzymatic glycation of free amino acids with elevated glucose via the Maillard reaction, which is commonly associated with chronic diseases and advanced glycation end-products.54 Our previous study also revealed that N-galactated amino acids represent pathognomonic markers of galactosemia in DBS extracts caused by deleterious accumulation of galactose.38 Although CF-related diabetes55 is not often diagnosed until in childhood/adolescence following progressive pancreatic fibrosis, earlier presentations have been reported in CF patients during infancy with recommendations to screen for impaired glucose tolerance earlier in life.56 For the first time, we identified a novel mixed oxidized glutathione disulfide conjugate with a N-glycated Gly (Glc-Gly) moiety when using MS/MS together with selective disulfide chemical reduction using excess DTT (Fig. 3) that was significantly elevated in asymptomatic CF infants relative to SN/healthy neonates similar to other N-glycated amino acids (e.g., Glc-Glu, Glc-Gln), including free Glc-Gly (Table 2).


Page 16 of 36

Page 17 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Additionally, CF infants were found to have much lower oxidized glutathione (GSSG), nicotinamide, and acetylcarnitine (C2), together with a depletion of (conditionally) essential amino acids in circulation, including modified amino acids/intermediates, such as 3methylhistidine and ornithine. All DBS extracts were found to have no detectable levels of reduced glutathione (GSH) despite its high abundance in erythrocytes suggesting that GSH is largely oxidized to GSSG during DBS specimen collection, transportation and storage. As a result, lower circulating levels of (total) glutathione in CF affected neonates is consistent with previous studies demonstrating that CFTR dysfunction in exocrine epithelial cells not only impacts chloride transport, but also nucleotide-regulated glutathione efflux.57,

58

This effect

contributes to oxidative/redox stress and reduced anti-mucolytic activity that plays a key role in CF pathophysiology.59,

60

For instance, impaired GSH efflux to the surface fluid coating of

airway epithelial cells lowers antioxidant capacity and promotes CF-related lung disease due to chronic inflammation and recurrent P. aeruginosa infections.61 Similarly, serum metabolomic studies of CF children compared to non-CF lung disease subjects demonstrated higher oxidative stress markers and lower medium-chain acylcarnitines suggestive of mitochondrial dysfunction.31 Although not detected by MSI-CE-MS due to insufficient concentration sensitivity when injecting nanolitres of sample on-capillary,38 NBS of a targeted panel of 44 biomarkers for various IEM using DI-MS/MS confirmed that several short- and medium-chain acylcarnitines, as well as amino acids were depleted in CF neonates as compared to healthy controls (Table S3). Recently, a plasma metabolomics study of the vitamin D status of adult CF patients relative to healthy subjects reported elevated glucose and lower amino acids reflective of a catabolic state that may be improved by high-dose vitamin D supplementation during acute pulmonary exacerbations.62 Similarly, low bioavailability of circulating Gln63 and Arg64 are associated with pulmonary disease in CF as a result of neutrophil-associated inflammation of lung tissue and nitric oxide deficiency in children, respectively. In order to elucidate CF-specific biomarkers in asymptomatic neonates, a comparison of the metabolic phenotype of true CF cases was compared to SP/non-CF neonates subsequently determined to be unaffected based on low sweat chloride test results (< 29 mM). Several amino acids were found to be significantly and consistently depleted (ranging from 10-30%) in CFaffected neonates relative to unaffected carriers and hypertrypsinogenemic cases as summarized 17 ACS Paragon Plus Environment


in Table 3. For example, Tyr, Thr, Ser, Pro, Gly, Asn, Arg, Ala, Gln and 3-MeHis were all found to be significantly decreased in CF neonates after adjustment for FDR (q < 0.05) relative to both SN/healthy and SP/non-CF cases (Table S1). These results indicate that CF neonates already suffer from maldigestion/malabsorption of dietary protein due to exocrine pancreatic insufficiency with lower prandial enzyme secretion.65 Indeed, earlier studies have reported altered Tyr metabolism was likely attributed to unabsorbed Tyr in the lumen of large intestine of CF children.66, 67 This is consistent with lower circulating Tyr in DBS extracts of CF neonates in this work given that excessive loss of fecal amino acids67 occurs in CF patients who are prone to malnutrition. Remarkably, Tyr is already included within a panel of amino acids used for NBS of phenylketonuria and tyrosinemia when using DI-MS/MS, but it is not applied for CF screening given the reliance on the two-tier IRT/DNA screening algorithm. In fact, Tyr and Gly were the two most significant amino acids depleted in CF as compared to SP/non-CF neonates based on DI-MS/MS data acquired from NSO, which was independently replicated by MSI-CE-MS using matching stored DBS cut-out specimens (Table S4). Importantly, excellent mutual agreement for Tyr quantification (Fig. 6S) confirms the lack of bias between two different MS platforms despite long-term DBS storage that may impact chemical stability when using metabolomics for biomarker discovery in retrospective studies. Importantly, this work also highlights the potential for facile translation of CF-specific biomarkers discovered by MSI-CE-MS when using existing MS/MS infrastructure already available at accredited NBS facilities. However, the targeted metabolite panel used for NBS by DI-MS/MS does not currently include several other amino acids demonstrated to be lower in CF neonates in our study, including Thr, Ser, Asn, Arg, Pro, Ala, Gln and 3-MeHis. Additionally, two other CF-specific biomarkers were OPA, and an unknown peptide elevated in CF that was the only metabolite directly correlated with IRT (ρ = 0.332, p = 4.55 E-4) as depicted in Fig. 4E. Indeed, OPA was tentatively identified (level 2) by high resolution MS/MS (Fig 1E),48 which is a glutathione analog and biomarker of oxidative stress due to hepatic glutathione depletion.68 As OPA is a chemically stable metabolite lacking a sulfhydryl moiety in DBS specimens, it may serve as a proxy for assessment of CFTR mediated impaired GSH efflux that is prevalent in CF.57 We thus hypothesize that glutathione and OPA may be co-transported by CFTR in exocrine epithelial tissue (e.g., lungs, intestines, pancreas) reflecting lower OPA levels measured in the circulation


Page 18 of 36

Page 19 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


of CF neonates. Indeed, both OPA and GSSG were previously reported to be depleted in airway epithelial cells of CF patients relative to non-CF donors representing an important mechanism in the pathogenesis of CF cells.32 The unknown trivalent ion was tentatively assigned as a decapentapeptide after interpretation of MS/MS spectra with product ions matching fragments containing Ile/Leu, and Val residues (Fig. S5). Unambiguous structural elucidation of the CFspecific peptide remains incomplete, but it may represent a by-product of trypsinogen activation due to its correlation with IRT reflecting underlying pancreatic insufficiency. In fact, small pancreatic duct blockage begins in utero, eventually leading to post-natal large duct obstruction and the subsequent secretion of proteins, including IRT, from the pancreas into the bloodstream in CF early in life.69 Further work is needed to sequence this peptide and define its exact biochemical role. Fig. 4D demonstrates that a ROC curve based on the measured ratio of OPA/unknown peptide may serve as a putative biomarker for discriminating true CF neonates from unaffected carriers and hypertrypsinogenemic cases. The present study is not without limitations. Samples were analyzed using a method optimized for the resolution and detection of hydrophilic/cationic metabolites using positive ionmode electrospray ionization. Indeed, most amino acids and acylcarnitines that serve as biomarkers for dozens of other genetic diseases in the population currently screened in NBS programs using multiplexed MS/MS technology with positive ion mode detection.48 Thus, metabolome coverage was limited since most acidic/lipophilic metabolites are optimally detected when using MSI-CE-MS under alkaline conditions with negative-ion mode detection. This was in fact critical to discover two exogenous anionic metabolites secreted at lower concentrations in the sweat of CF infants that was indicative of CFTR-related paraoxanase deficiency, namely pilocarpic acid and mono(2-ethylhexyl)phthalic acid.39 Also, current efforts are underway to extend MSI-CE-MS when using non-aqueous buffer conditions for multiplexed analysis of fatty acids and other lipid classes from a single DBS punch. This study also lacked an independent holdout or validation set that is necessary for rigorous biomarker validation from differentiating metabolites.69 Given the overall birthrate and incidence of CF in the province of Ontario, obtaining an adequate number of CF cases was a practical constraint. Nevertheless, independent data replication confirmed systemic amino acid depletion in CF neonates when using two validated MS platforms performed on recently collected (DI-MS/MS) and stored (MSI-CE-MS)



DBS specimens. Follow-up work is planned for validating lead candidate biomarkers within a larger cohort from multiple NBS centers as a way to increase study power. Additionally, a prospective study is needed for long-term assessment of the clinical utility and cost savings of CF-specific biomarkers measured by DI-MS/MS as compared to the current IRT/DNA screening algorithms. This work is urgently required to reduce the high rate of false positives from newborn CF screening and unnecessary follow-up molecular diagnostic (genetic) and physiological (sweat chloride) testing that contribute to higher health care expenditures and potential psychosocial harms.70 Additionally, metabolic phenotyping of CF neonates provides deeper insights into the functional assessment of CFTR mutations of unknown or variable consequence with the advent of next generation sequencing.71 Although once considered a fatal genetic disorder in children due to severe malnutrition, CF is now a manageable chronic disease with the advent of newborn screening, early therapeutic interventions and access to health care resources as reflected by a median age of survival of over 48 years for CF patients in Canada.72 Conclusion In summary, this work presents the first non-targeted metabolomics study for pre-symptomatic detection of CF from retrospective neonatal DBS specimens. A fully validated metabolomics workflow was applied with stringent quality control measures that allowed for effective batch correction of data with excellent technical precision. Preliminary data suggest that CF neonates suffer from protein maldigestion/malabsorption resulting from exocrine pancreatic insufficiency as reflected by lower concentrations of several circulating amino acids (notably Tyr, Ser, and Thr), whereas decreased levels of OPA is indicative of impaired glutathione efflux due to CFTR dysfunction. Prospective studies are still needed to clinically validate the utility of CF-specific biomarkers for NBS that can be measured using existing DI-MS/MS infrastructure at incremental costs as a way to improve positive predictive value. Reducing costly and unnecessary testing will alleviate familial anxiety and reduce health care expenditures by focusing resources to those truly affected by CF. Metabolic phenotyping of neonates is anticipated to help resolve other diagnostic dilemmas in CF, including false-negatives who are not detected by IRT screens, screen-positives having CFTR mutations with unknown or variable consequences, and presumptive CF cases with an inconclusive diagnosis and poorly understood prognosis.


Page 20 of 36

Page 21 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Acknowledgments P.B.M. wishes to acknowledge funding support from the Cystic Fibrosis Canada, Natural Sciences and Engineering Research Council of Canada, Canada Foundation for Innovation and Genome Canada. Further thanks are directed to Dr. Marcus Kim at Agilent Technologies Inc., the Faculties of Science and Health Sciences at McMaster University, and the CDC’s Newborn Screening Quality Assurance Program. SUPPORTING INFORMATION The following supporting information is available free of charge on the ACS publications website http://pubs.acs.org. Supporting methods description, including cohort selection and study design; IRT and CFTR mutational analysis for CF newborn screening; Metabolite screening by stable-isotope dilution FIA-MS/MS. Supporting data files (xlsx and xml) are also included, including original metabolomics data matrix and batched-corrected data matrix with QCs, as well as MS/MS spectra for two tentatively identified ions, namely divalent glutathione disulfide derivative and unknown polypeptide. Results from a Kruskal-Wallis test on batch-corrected data between CFpositive, SP and screen-negative cases, Table S1. Metabolite panel analyzed from DBS specimens using stable-isotope dilution DI-MS/MS at NSO, Table S2. Mann-Whitney U test for ranking metabolites from DBS measured by DI-MS/MS at NSO that differentiate CF relative to SN/healthy neonates, Table S3. Mann-Whitney U test for ranking metabolites from DBS measured by DI-MS/MS at NSO that differentiate CF from SP/non-CF neonates, Table S4. A QC-based batch correction algorithm applied on MS-based metabolomics data when adjusting for long-term instrumental drift, Figure S1. Metabolic signatures from DBS extracts associated with CF as compared to screen-negative/healthy neonates, Figure S2. High resolution MS/MS for a series of N-glycated amino acids acids that were significantly elevated in DBS extracts of CF neonates as compared to SN/healthy controls, Figure S3. Metabolic signatures from DBS extracts associated with CF neonates as compared to unaffected screen-positive carriers, Figure S4. A mirror plot comparing MS/MS spectra for ophthalmic acid (OPA) measured in DBS extract and authentic chemical standard for unambiguous identification, Figure S5. Structural elucidation of an unknown trivalent peptide using high resolution MS and MS/MS, Figure S6. An inter-laboratory method comparison for quantification of tyrosine (Tyr) from matching DBS specimens analyzed independently by DI-MS/MS and MSI-CE-MS, Figure S7. REFERENCES (1) Tsui, L. C.; Dorfman, R. The Cystic Fibrosis Gene: A Molecular Genetic Perspective. Cold Spring Harb. Perspect. Med. 2013, 3, a009472. (2)

Sosnay, P.R.; Siklosi, K. R.; van Goor, F.; Kaniecki K.; Yu, H.; Sharma, N.; Raamalho, A. S.; Amaral, M. D.; Dorfman, R.; Zielenski, J. et al. Defining the Disease Liability of Variants in the Cystic Fibrosis Transmembrane Conductance Regulator Gene. Nature Genetics 2013, 45, 1160−1167. 21 ACS Paragon Plus Environment


(3)

Li, W.; Soave, D. Miller, M. R.; Kennan, K.; Lin F.; Gong, J.; Chiang, T.; Stephenson, A. L.; Durie, P. Rommens, J.; Sun, L.; Strug, L. J. Unraveling the Complex Genetic Model for Cystic Fibrosis: Pleiotropic Effects of Modifier Genes On Early Cystic Fibrosisrelated Morbidities. Hum. Genet. 2014, 133, 151−161.

(4)

Dijk, F. N.; McKay, K.; Barzi, F.; Gaskin, K. J.; Fitzgerald., D. A. Improved Survival in Cystic Fibrosis Patients Diagnosed by Newborn Screening Compared to a Historical Cohort from the Same Centre. Arch. Dis. Child. 2011, 96, 1118−1123.

(5)

Farrell, P. M.; Lai, H. J.; Li, Z.; Kosorok, M. R.; Laxova, A.; Green, C. G.; Collins, J.; Hoffman, G.; Laessig, R.; Rock, M. J.; Splaingard, M. L. Evidence on Improved Outcomes with Early Diagnosis of Cystic Fibrosis Through Neonatal Screening: Enough is Enough! J. Pediatr. 2005, 147, S30−S36.

(6)

Sims, E. J.; Clark, A.; McCormick, J.; Mehta, G.; Connett, G.; Mehta, A. Cystic Fibrosis Diagnosed After 2 Months of Age Leads to Worse Outcomes and Requires More Therapy. Pediatrics 2007, 119, 19-28.

(7)

Grosse, S. D. Showing Value in Newborn Screening: Challenges in Quantifying the Effectiveness and Cost-effectiveness of Early Detection of PKU and Cystic Fibrosis. Healthcare 2015 3, 1133−1157.

(8)

de Boeck, K.; Munck, A.; Walker, S.; Faro, A.; Hiatt, P.; Gilmartin, G.; Higgins, M. Efficacy and Safety of Ivacaftor in Patients with Cystic Fibrosis and a Non-G551D Gating Mutation. J. Cyst. Fibros. 2014 13, 674−680.

(9)

Wainwright, C. E.; Elborn, J. S.; Ramsey, B. W.; Marigowda, G.; Huang X.; Cipolli, M.; Colombo, C.; Davies, J. C.; de Boeck, K.; Flume, P. A. et al. Lumacaftor-Ivacaftor in Patients with Cystic Fibrosis Homozygous for Phe508del CFTR. New Engl. J. Med. 2015 373, 1783 –1784.

(10)

Farrell, P. M. White, T. B.; Ren, C. L.; Hempstead, S. E.; Accurso, F.; Derichs, N.; Howenstine, M.; McColley, S. A.; Rock, M.; Rosenfeld, M. et al. Diagnosis of Cystic Fibrosis: Consensus Guidelines from the Cystic Fibrosis Foundation. J. Pediatr. 2017, 181, S4–S15.

(11)

Massie, R. J. H.; Curnow, L.; Glazner, J.; Armstrong, D. S.; Francis, I. Lessons Learned from 20 Years of Newborn Screening for Cystic Fibrosis. Med. J. Aust. 2011, 196, 67– 70.

(12)

Baker, M. W.; Groose, M.; Hoffman, G.; Rock, M.; Levy, H.; Farrell, P. M. Optimal DNA Tier for the IRT/DNA Algorithm Determined by CFTR Mutation Results Over 14 Years of Newborn Screening. J. Cyst. Fibros. 2011, 10, 278–281.

(13)

Crossley, J. R.; Elliott, R. B.; Smith, P. A. Dried-blood Spot Screening For Cystic Fibrosis In The Newborn. Lancet 1979, 313, 472–474.

(14)

Ryley, H. C.; Robinson, P. G.; Yamashiro, Y.; Bradley, D. M. Assay of Serum Immunoreactive Trypsin in Dried Blood Spots and the Early Detection of Cystic Fibrosis. J. Clin. Pathol. 1981, 34, 906–910.

(15)

Ranieri, E.; Ryall, R. G.; Morris, C. P.; Nelson, P. V.; Carey, W. F.; Pollard, A. C.; 22 ACS Paragon Plus Environment

Page 22 of 36

Page 23 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Robertson, E. F. Neonatal Screening Strategy for Cystic Fibrosis Using Immunoreactive Trypsinogen and Direct Gene Analysis. BMJ 1991, 302, 1237–1240. (16)

Levy, H.; Farrell, P. M. New Challenges in the Diagnosis and Management of Cystic Fibrosis. J. Pediatr. 2015, 166, 1337–1341.

(17)

Kay, D. M. Maloney, B.; Hamel, R.; Pearce, M.; DeMartino, L.; McMahon, R.; McGrath, E.; Krein, L.; Vogel, B.; Saavedra-Matiz, C. A. et al. Screening for Cystic Fibrosis in New York State: Considerations for Algorithm Improvements. Eur. J. Pediatr. 2016, 175, 181–193.

(18)

Frazier, D. M.; Millington, D. S.; McCandless, S. E.; Koeberl, D. D.; Weavil, S. D.; Chaing, S. H.; Muenzer, J. The Tandem Mass Spectrometry Newborn Screening Experience in North Carolina: 1997-2005. J. Inherit. Metab. Dis. 2006, 29, 76–85.

(19)

Schrijver, I.; Pique, L.; Graham, S.; Pearl, M.; Cherry, A,; Kharrazi, M. The Spectrum of CFTR Variants in Nonwhite Cystic Fibrosis Patients: Implications for Molecular Diagnostic Testing. J. Mol. Diagn. 2016, 18, 39-50.

(20)

Sontag, M. K.; Lee, R.; Wright, D.; Freedenberg, D.; Sagel, S. D. Improving the Sensitivity and Positive Predictive Value in a Cystic Fibrosis Newborn Screening Program Using a Repeat Immunoreactive Trypsinogen and Genetic Analysis. J. Pediatr. 2016, 175, 150–158.

(21)

Sommerburg, O.; Krulisova, V.; Hammermann, J.; Lindner, M.; Stahl, M.; Muckenthaler, M.; Kohlmueller, D.; Happich, M.; Kulozik, A. E.; Votava, F. et al. Comparison of Different IRT-PAP Protocols to Screen Newborns for Cystic Fibrosis in Three Central European Populations. J. Cyst. Fibros. 2014, 13, 15–23.

(22)

Quinton, P. M. Chloride Impermeability in Cystic Fibrosis. Nature 1983, 301, 421– 422.

(23)

Collie, J. T. B.; Massie, R. J.; Jones, O. A. H.; Legrys, V. A.; Greaves, R. F. Sixty-five Years Since the New York Heat Wave: Advances in Sweat Testing for Cystic Fibrosis. Pediatr. Pulmonol. 2014, 49, 106–117.

(24)

Kirk, J. M. A Continuing Role for Sweat Testing in an Era of Newborn Screening for Cystic Fibrosis. Clin. Biochem. 2011, 44, 487–488.

(25)

Ooi, C. Y.; Castellani, C.; Keenan, K.; Avolio, J.; Volpi, S.; Boland, M.; Kovesi, T.; Bjornson, C.; Chilvers, M. A.; Morgan, L. et al. Inconclusive Diagnosis of Cystic Fibrosis After Newborn Screening. Pediatrics 2015, 135, e1377–e1385.

(26)

Beger, R. D.; Dunn, W.; Schmidt, M. A.; Gross, S. S.; Kirwan, J. A.; Cascante, M; Brennan, L.; Wishart, D. S.; Oresic, M.; Hankemeier, T. et al. Metabolomics Enables Precision Medicine: “A White Paper, Community Perspective” Metabolomics 2016, 12, 149.

(27)

Newgard, C. B. Metabolomics and Metabolic Diseases: Where do We Stand? Cell Metabol. 2017, 25, 43–56.

(28)

Johnson, C. H.; Ivanisevic, J.; Siuzdak, G. Metabolomics: Beyond Biomarkers and 23 ACS Paragon Plus Environment


Towards Mechanisms. Nat. Rev. Mol. Cell Biol. 2016, 17, 451–459. (29)

Raamsdonk, L. M.; Teusink, B.; Broadhurst, D.; Zhang, N.; Hayes, A. et al. A Functional Genomics Strategy That Uses Metabolome Data to Reveal the Phenotype of Silent Mutations. Nat Biotechnol 2001, 19, 45–50.

(30)

Muhlebach, M. S.; Sha,W. Lessons Learned From Metabolomics in Cystic Fibrosis. Mol. Cell. Pediatr. 2015, 2, 1–7.

(31)

Joseloff, E.; Sha, W. Bell, S. C.; Wetmore, D. R.; Lawton, K. A.; Milburn, M. V. Ryals, J. A.; Guo, L.; Muhlebach, M. S. Serum Metabolomics Indicate Altered Cellular Energy Metabolism in Children with Cystic Fibrosis. Pediatr. Pulmonol. 2014, 49, 463–472.

(32)

Wetmore, D. R.; Joseloff, E.; Pilewski, J.; Lee, D. P.; Lawton, K. A.; Mitchell, M. W.; Milburn, M. V.; Ryals, J. A.; Guo, L. Metabolomic Profiling Reveals Biochemical Pathways and Biomarkers Associated with Pathogenesis in Cystic Fibrosis Cells. J. Biol. Chem. 2010, 285, 30516–30522.

(33)

Robroeks, C. M.; van Berkel, J. J.; Dallinga, J. W.; Jobsis, Q.; Zimmermann, L. J.; Hendriks, H. J.; Wouters, M. F.; van der Grinten, C. P.; van de Kant, K. D.; van Schooten, F. J.; Dompeling, E. Metabolomics of Volatile Organic Compounds in Cystic Fibrosis Patients and Controls. Pediatr. Res. 2010, 68, 75–80.

(34)

Wolak, J. E.; Esther, C. R.; O’Connell, T. M. Metabolomic Analysis of Bronchoalveolar Lavage Fluid From Cystic Fibrosis Patients. Biomarkers 2009, 14, 55–60.

(35)

Esther, C. R.; Coakley, R. D.; Henderson, A. G.; Zhou, Y. H.; Wright, F. A.; Boucher, R. C. Metabolomic Evaluation of Neutrophilic Airway Inflammation in Cystic Fibrosis. Chest 2015, 148, 507–515.

(36)

Esther, C. R.; Turkovic, L.; Rosenow, T.; Muhlebach, M. S.; Boucher, R. C.; Ranganathan, S.; Stick, S. M. Metabolomic Biomarkers Predictive of Early Structural Lung Disease in Cystic Fibrosis. Eur. Respir. J. 2016, 48, 1612–1621.

(37)

Quinn, R. A.; Lim, Y. W.; Mak, T. D.; Whiteson, K.; Furlan, M.; Conrad, D.; Rohwer, F.; Dorrestein, P. Metabolomics of Pulmonary Exacerbations Reveals the Personalized Nature of Cystic Fibrosis Disease. Peer J. 2016, 4, 1–21.

(38)

DiBattista, A.; McIntosh, N.; Lamoureux, M.; Al-Dirbashi, O. Y.; Chakraborty, P.; Britz-McKibbin, P. Temporal Signal Pattern Recognition in Mass Spectrometry: A Method for Rapid Identification and Accurate Quantification of Biomarkers for Inborn Errors of Metabolism with Quality Assurance. Anal. Chem. 2017, 89, 8112–8121.

(39)

Macedo, A. N.; Mathiaparanam, S.; Brick, L.; Keenan, K.; Gonska, T.; Pedder, L.; Hill, S.; Britz-McKibbin, P. The Sweat Metabolome of Screen-Positive Cystic Fibrosis Infants: Revealing Mechanisms beyond Impaired Chloride Transport. ACS Cent. Sci. 2017, 3, 904–913.

(40)

McHugh, D.; Cameron, C. A.; Abdenur, J. E.; Abdulrahman, M.; Adair, O.; Al Nuaimi, S. A.; Ahlman, H.; Allen, J. J.; Antonozzi, I. Archer, S. et al. Clinical Validation of Cutoff Target Ranges in Newborn Screening of Metabolic Disorders by Tandem Mass Spectrometry: A Worldwide Collaborative Project. Genet. Med. 2011, 13, 230–254. 24 ACS Paragon Plus Environment

Page 24 of 36

Page 25 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(41)

Kuehnbaum, N. L.; Kormendi, A.; Britz-McKibbin, P. Multisegment Injection-Capillary Electrophoresis-Mass Spectrometry: A High Throughput Platform for Metabolomics with High Data Fidelity. Anal. Chem. 2013, 85, 10664–10669.

(42)

Mak, D. Y. F.; Sykes, J.; Stephenson, A. L.; Lands, L. C. The Benefits of Newborn Screening for Cystic Fibrosis: The Canadian Experience. J. Cyst. Fibros. 2016, 15, 302– 308.

(43)

Hayeems, R. Z.; Miller, F. A.; Vermeulen, M. Potter, B. K.; Chakrabory, P.; Davies, C.; Carroll, J. C.; Ratjen, F.; Guttmann, A. False-Positives Newborn Screening for Cystic Fibrosis and Health Care Use. Pediatrics 2017, 140, e20170604.

(44)

Wehrens, R.; Hageman, J. A.; van Eeuwijk, F.; Kooke, R.; Flood, P. J. Wijnker, E.; Keurentjes, J. J.; Lommen, A.; van Eekelen, H. D, Hall, R. D.; Mumm, R.; de Vos, R. C. Improved Batch Correction in Untargeted MS-based Metabolomics. Metabolomics 2016, 12, 1–12.

(45)

Kamleh, M. A.; Ebbels, T. M. D.; Spagou, K.; Masson, P.; Want, E. J. Optimizing the Use of Quality Control Samples for Signal Drift Correction in Large-scale Urine Metabolic Profiling Studies. Anal. Chem. 2012, 84, 2670–2677.

(46)

Salek, R.M; Steinbeck, C.; Viant, M.R.; Goodacre, R.; Dunn, W.B. The Role of Reporting Standards for Metabolite Annotation and Identification in Metabolomic Studies. Gigascience 2013, 2, 13.

(47)

Chong, J.; Soufan, O.; Li, C.; Caraus, I.; Li, S.; Bourque, G.; Wishart, D. S.; Xia, J. MetaboAnalyst 4.0: Towards More Transparent and Integrative Metabolomics Analysis. Nucl. Acids Res. 2018, 46, W486–W494.

(48)

Soga, T.; Baran, R.; Suematsu, M.; Ueno, Y.; Ikeda, S.; Sakurakawa, T.; Kakazu, Y.; Ishikawa, T.; Robert, M.; Nishoka, T., Tomita, M. Differential Metabolomics Reveals Ophthalmic Acid as an Oxidative Stress Biomarker Indicating Hepatic Glutathione Consumption. J. Biol. Chem. 2006, 281, 16768–16776.

(49)

Chace, D. H. A Mass Spectrometry in Newborn and Metabolic Screening: Historical Perspective and Future Directions. J. Mass. Spectrom. 2009, 44, 163-170.

(50)

Sims, E. J.; Clark, A.; McCormick, J.; Mehta, G.; Connett, G.; Mehta, A. Cystic Fibrosis Diagnosed After 2 Months of Age Leads to Worse Outcomes and Requires More Therapy. Pediatrics 2007, 119, 19–28.

(51)

Dijk, F. N.; McKay, K.; Barzi, F.; Gaskin, K. J. ; Fitzgerald, D. A. Improved Survival in Cystic Fibrosis Patients Diagnosed by Newborn Screening Compared to a Historical Cohort From the Same Centre. Arch. Dis. Child. 2011, 96, 1118–1123.

(52)

Schmidt, J. L.; Castellanos-Brown, K.; Childress, S.; Bonhomme, N.; Oktay, J. S.; Terry, S. F.; Kyler, P.; Davidoff, A.; Greene, C. The Impact of False-positive Newborn 25 ACS Paragon Plus Environment


Screening Results on Families: A Qualitative Study. Genet. Med. 2012, 14, 76–80. (53)

Currier, R. J.; Scortino, S.; Liu, R.; Bishop, T.; Koupaei, R. A.; Feuchtbaum, L. Genomics Sequecning in Cystic Fibrosis Newborn Screening: What Works Best, Twotier Predefined CFTR Mutation Panels or Second-tier CFTR Panel Followed by Thirdtier Sequencing? Gen. Med. 2017, 19, 1159–1163.

(54)

Zhang, Q.; Ames, J. M.; Smith, R. D.; Baynes, J. W.; Metz, T. O. A Perspective on the Maillard Reaction and the Analysis of Protein Glycation by Mass Spectrometry: Probing the Pathogenesis of Chronic Disease. J. Proteome Res. 2008, 8, 754–769.

(55)

Ntimbane, T,; Comte, B.; Mailhot, G.; Berthiaume, Y.; Poitout, V.; Prentki, M.; RabasaLhoret, R.; Levy, E. Cystic Fibrosis-related Diabetes: From CFTR Dysfunction to Oxidative Stress. Clin. Biochem. Rev. 2009, 30, 153–177.

(56)

Gelfand, I. M.; Eugster, E. A.; Haddad, N. G. Infancy-Onset Cystic Fibrosis-Related Diabetes. Diabetes Care 2005, 28, 2693–2594.

(57)

Linsdell P.; Hanrahan, J.W. Glutathione permeability of CFTR. Am. J. Physiol. 1998, 275, C323–C326.

(58)

Hudson, V. M. Rethinking Cystic Fibrosis Pathology: The Critical Role of Abnormal Reduced Glutathione (GSH) Transport Caused by CFTR Mutation, Free Radic. Biol. Med. 2001, 30, 1440–1461.

(59)

Kogan, I.; Ramjeesingh, M.; Li, C.; Kidd, J. F.; Wang, Y.; Leslie, E. M.; Cole, S. P. C. ; Bear, C. E. CFTR Directly Mediates Nucleotide-regulated Glutathione Flux. EMBO J. 2003, 22, 1981–1989.

(60)

Hudson, V. M. New Insights into the Pathogenesis of Cystic Fibrosis: Pivotal Role of Glutathione System Dysfunction and Implications for Therapy. Treat. Respir. Med. 2004, 3, 353–363.

(61)

Kariya, C.; Leitner, H.; Min, E.; van Heeckeren, C.; van Heeckeren, A.; Day, B. J. A Role for CFTR in the Elevation of Glutathione Levels in the Lung by Oral Glutathione Administration. Am. J. Physiol. Lung Cell. Mol. Physiol. 2007, 292, L1590–L1597.

(62)

Alvarez, J. A. ; Chong, E. Y. ; Walker, D. I.; Chandler, J. D.; Michalski, E. S. ; Grossmann, R. E. ; Li, S.; Frediani, J. K.; Tirouvanziam, R. ; et al. Plasma Metabolomics in Adults with Cystic Fibrosis During a Pulmonary Exacerbation: A Pilot Randomized Study of High-dose Vitamin D3 Administration. Metabolism 2017, 70, 31–41.

(63)

D’Eufemia, P.; Finocchiaro, R.; Celli, M.; Tote, J.; Ferrucci, V.; Zambrano, A.; Troiani, P.; Quattrucci, S. Neutrophil Glutamine Deficiency in Relation to Genotype in Children with Cystic Fibrosis, Pediatr. Res. 2006, 59, 13–16.

(64)

Engelen, M. P.; Com, G.; Luiking, Y. C.; Deutz, N. E. Stimulated Nitric Oxide Production and Arginine Deficiency in Children with Cystic Fibrosis with Nutritional Failure. J. Pediatr. 2013, 163, 369–376. 26 ACS Paragon Plus Environment

Page 26 of 36

Page 27 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(65)

Li, L.; Somerset, S. Digestive System Dysfunction in Cystic Fibrosis: Challenges for Nutrition Therapy, Dig. Liver Dis. 2014, 46, 865–874.

(66)

Robinson, R. Tyrosine Metabolism in Cystic Fibrosis. Clin. Chim. Acta 1966, 14, 166– 170.

(67)

Gibbons, I. S.; Seakins, J. W.; Ersser, R. S. Tyrosine Metabolism and Faecal Amino Acids in Cystic Fibrosis of the Pancreas. Lancet 1967, 289, 877–878.

(68)

Dello, S. A.; Neis, E. P.; de Jong, M. C.; van Eijk, H. M.; Kicken, C. H.; Olde Damink, S. W.; Dejong, C. H. Systematic Review of Opthalmate as a Novel Biomarker of Hepatic Glutathione Depletion. Clin. Nutr. 2013, 32, 325–330.

(69)

Koulman, A.; Lane, G. A.; Harrison, S. J.; Volmer, D. A. From Differetniating Metabolites to Biomarkers. Anal. Bioanal. Chem. 2009, 394, 663–670.

(70)

Hayeems, R. Z.; Miller, F. A.; Barg, C. J.; Bombard, Y.; Kerr, E. Parent Experience With False-Positive Newborn Screening Results for Cystic Fibrosis. Pediatrics 2016, 138, 1– 10.

(71)

Lefterova, M. I.; Shen, P.; Odegaard, J. I. ; Fung, E.; Chiang, Peng, G.; Davis, R. W.; Wang, W.; Kharrazi, M.; Schrijver, I.; Scharfe, C. Next-Generation Molecular Testing of Newborn Dried Blood Spots for Cystic Fibrosis. J. Mol. Diagnostics 2016, 18, 267–282.

(72)

Stephenson, A. L.; Sykes, J.; Stanojevic, S.; Quon, B. S.; Marshall, B. C.; Petren, K.; Ostrenga, J.; Fink, A. K.; Elbert, A.; Goss, C. H. Survival Comparison of Patients with Cystic Fibrosis in Canada and United States: A Population-based Cohort Study. Ann. Intern. Med. 2017, 166, 537-546.



Page 28 of 36

Table 1. Study cohort for metabolomics study of presumptive screen-positive (CF; non-CF) and screennegative (SN/healthy) neonates based on a two-tiered IRT/DNA algorithm followed by sweat chloride test for confirmatory CF diagnosis. Variablea True CF Carriers HypertrypScreen negative (n=36) (n=37) sinogenemic (n=44) (n=35) Sex Female (#) Male (#) Birth weight (g) mean ± SD IRT (ng/mL) median ± IQR Sweat Cl- (mmol/L)b median ± IQR CFTR genotypec 0 mutations 1 mutation: ∆F508/1 mutation: other/2 mutations: ∆F508/∆F508 2 mutations: ∆F508/other 2 mutations: other/other Pancreatic Statusd Sufficient Insufficient

18 18 3156 ± 503

18 19 3465 ± 435

20 15 3465 ± 357

22 22 3442 ± 440

179 ± 141

61 ± 33

128 ± 63

17.3 ± 9.3

89 ± 10

13 ± 5

11 ± 1.4

n/a

---24

-24 13 --

35 ----

44 ----

10 2

---

---

---

1 35

n/a

n/a

n/a

a All

variables were tested for normality using a Shapiro-Wilks test and only birth weight was significant (p < 0.05) pilocarpine-stimulated iontophoresis using macrobore collectors were used for confirmatory diagnosis of all presumptive CF cases at regional CF clinics in the province of Ontario. c NSO uses a 39 CFTR mutation panel, including 4 variants in a two-tier IRT/DNA algorithm for CF detection. d Fecal elastase-1 measurements performed to assess pancreas status of confirmed CF cases with high sweat chloride at regional CF clinics in the province of Ontario using a cut-off of 100 µg/g for pancreatic insufficiency. b Standardized


Page 29 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Table 2. Mann-Whitney U test to rank metabolites from DBS extracts that were significantly different (q < 0.05) in CF (n = 36) relative to screen-negative (n = 40) neonates by MSI-CE-MS. Mass FDR Effect FCc Metabolite ID m/z:RMT p-value Error q-value Size (CF/Neg) (ppm) Glc-Glutamic acidb Glc-Glutamineb Glc-Leucineb Glc-GSSGb Glutamine Threonine Nicotinamide Acetylcarnitine Oxidized Glutathione Glc-Glycineb Tyrosine Histidine Glycine Ornithine 3-Methylhistidine Serine Unknown M+2H2+ Asparagine Methionine Lysine Homoarginine* Proline Unknown Unknown M+3H3+ Ophthalmic Acid Arginine Glutamic Acid Kynurenine Unknown Valine Trimethyllysine* Unknown

310.115:1.430

3.39

9.05 E-13a

6.33 E-11

0.774

3.15

309.129:1.421 294.155:1.283 388.109:1.300 147.076:0.960 120.066:0.930

-1.49 -2.98 -0.81 -1.39 0.24

1.09 E-09a 3.56 E-08a 4.35 E-08a 4.65 E-08a 8.38 E-08a

3.80 E-08 6.51 E-07 6.51 E-07 6.51 E-07 9.78 E-07

0.683 0.628 0.624 0.623 0.612

2.65 2.90 1.94 0.72 0.71

123.055:0.650 204.123:0.791

0.31 -1.05

2.10 E-06a 2.91 E-06a

2.15 E-05 2.26 E-05

0.550 0.542

1.38 0.68

307.083:1.109

-2.92

2.91 E-06a

2.26 E-05

0.542

0.73

238.092:1.175 182.081:1.015 156.077:0.630 76.039:0.720 133.097:0.582 170.092:0.648 106.050:0.880 521.798:0.970 133.061:0.930 150.058:0.941 147.113:0.585 189.135:0.612 116.071:0.952 162.076:0.795 438.258:0.833 290.135:1.225 175.119:0.608 148.060:0.976 209.092:0.919 161.129:0.603 118.086:0.868 189.160:0.609 252.163:0.582

1.14 -0.60 -0.58 0.13 0.18 -0.17 0.45 1.78 0.82 -0.1 -0.74 0.58 0.25 -0.06 -0.03 0.41 -0.14 -0.96 0.80 1.83 0.93 1.79 -1.00

3.61 E-06a 9.67 E-06a 1.67 E-05a 1.01 E-04a 1.05 E-04a 2.25 E-04 a 2.45 E-04 a 5.61 E-04 a 8.49 E-04 1.02 E-03 1.06 E-03 1.06 E-03 1.10 E-03 1.31 E-03 6.80 E-03 8.63 E-03 1.52 E-02 1.57 E-02 1.65 E-02 1.75 E-02 2.00 E-02 2.16 E-02 2.22 E-02

2.52 E-05 6.15 E-05 9.74 E-05 5.27 E-04 5.27 E-04 1.05 E-03 1.07 E-03 2.31 E-03 3.03 E-03 3.49 E-03 3.49 E-03 3.49 E-03 3.49 E-03 4.00 E-03 1.98 E-02 2.42 E-02 4.06 E-02 4.06 E-02 4.14 E-02 4.22 E-02 4.66 E-02 4.85 E-02 4.85 E-02

0.538 0.516 0.503 0.458 0.456 0.435 0.433 0.409 0.396 0.390 0.389 0.389 0.388 0.382 0.324 0.314 0.291 0.290 0.288 0.285 0.279 0.276 0.275

1.90 0.70 0.64 0.83 0.70 0.74 0.75 1.26 0.84 0.84 0.82 0.71 0.88 1.27 1.18 0.84 0.68 0.81 0.84 0.81 0.92 0.90 0.90

a These

metabolites were significant after a Bonferroni correction (p < 0.000725) based on accurate mass match and MS/MS fragmentation patterns38 c Median fold-change (FC) was calculated as the ratio of CF to SN/healthy neonates * Tentative identification based on accurate mass match from Human Metabolome Database, whereas four unknown ions have most likely molecular formulae of C44H83N9O19 (m/z 521.798), C6H15NO6 (m/z 162.076), C59H113N3O28 (m/z 438.258), and C22H42N6O7 (m/z 252.163) b Identification



Page 30 of 36

Table 3. Mann-Whitney U test to rank metabolites from DBS extracts significantly different (q < 0.05) in CF (n = 36) relative to screen-positive yet unaffected (non-CF) neonates (n = 72). Metabolite ID

m/z:RMT

Mass Error

p-value

(ppm) c

FDR q-value

Effect Size

FC b (CF/ non-CF)

a 182.081:1.015 -0.60 6.03 E-03 0.472 0.74 Tyrosine 6.14 E-05 c a 290.135:1.225 0.41 2.25 E-03 0.471 0.80 Ophthalmic Acid 6.52 E-05 c a 106.050:0.880 0.45 2.32 E-03 0.454 0.76 Serine 1.20 E-04 c a 120.066:0.930 0.24 2.32 E-03 0.451 0.80 Threonine 1.35 E-04 c a 116.071:0.952 0.25 6.30 E-03 0.416 0.92 Proline 4.57 E-04 Alanine 90.055:0.782 0.32 1.57 E-03 1.58 E-02 0.377 0.84 c 76.039:0.720 0.13 1.61 E-03 1.58 E-02 0.376 0.84 Glycine c 147.076:0.960 -1.39 2.02 E-03 1.70 E-02 0.368 0.84 Glutamine c 170.092:0.648 -0.17 2.22 E-03 1.70 E-02 0.365 0.84 3-Methylhistidine 2+* 179.129:0.651 3.99 2.60 E-03 1.79 E-02 0.359 0.87 Unknown M+2H Arginine 175.119:0.608 -0.14 3.54 E-03 2.22 E-02 0.348 0.70 3+ c* 438.258:0.833 -0.03 3.58 E-03 2.22 E-02 0.345 1.17 Unknown M+3H c 133.061:0.930 0.82 4.20 E-03 2.23 E-02 0.342 0.91 Asparagine Homoarginine* 189.135:0.612 0.58 5.29 E-03 2.61 E-02 0.333 0.82 Aminoisobutyric acid* 104.071:0.834 -0.46 6.37 E-03 2.93 E-02 0.326 0.86 Deoxycarnitine 146.118:0.700 0.03 1.13 E-02 4.87 E-02 0.304 0.90 a These metabolites were significant after a Bonferroni correction (p < 0.000725). b Median fold-change (FC) is calculated as the ratio of CF relative to SP/non-CF neonate. c These metabolites are also significantly different between CF infants and SN/healthy controls. * Tentative identification based on accurate mass match in Human Metabolome Database, whereas two unknown ions have most likely molecular formulae of C17H32N4O4 (m/z 179.129) and C59H113N3O28 (m/z 438.258)


Page 31 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 1. (A) A representative data workflow when using MSI-CE-MS with temporal signal pattern recognition based on a randomized injection series of three pairs of DBS extracts together with a pooled QC with each run. (B) A schematic depicting the 7-sample serial injection used in MSI-CE-MS (e.g., shown for run 1), where each pair of DBS extract is injected and encoded by a dilution pattern to facilitate peak assignment with quality assurance. (C) A series of extracted ion electropherograms for a batch of runs by MSI-CE-MS for an unknown ion annotated by its characteristic m/z:RMT that was selected after a dilution trend filter and QC check. (D) A full-scan high resolution mass spectrum for the single charged protonated molecule (MH+) with low mass error (< 1 ppm) and most likely molecular formula. (E) Structural elucidation of unknown ion after collision-induced dissociation (CID) of precursor ion resulting in formation of three diagnostic product ions that was consistent with ophthalmic acid (OPA).



Figure 2. (A) A PCA 2D score plot provides an overview of the metabolic phenotype of neonates in terms of their biological variance (CV = 34%) relative to technical variance (CV = 8%) measured for 70 polar metabolites from DBS extracts consistently detected in both screen-positive CF and screen-negative controls (> 75% frequency; CV < 40% in QC samples). (B) 32 different metabolites were differentially expressed in CF infants with high sweat chloride (n=36) as compared to healthy/screen-negative controls (n=44), including elevations in N-glycated amino acids, oxidized glutathione and nicotinamide, and lower amino acids and acetylcarnitine in circulation. (C) Excellent classification of CF neonates (n=36) from SN/healthy controls (n=44) was demonstrated by single or ratiometric biomarkers of CF when using ROC curves as reflected by an AUC > 0.92.


Page 32 of 36

Page 33 of 36 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 3. Structural elucidation of an unknown modified glutathione analog in DBS extracts based on (A) its most likely molecular formula for the divalent protonated molecule (MH22+) based on its accurate mass, charge state and isotopic pattern, (B) specific chemical reactivity to excess DTT, confirming it as a mixed oxidized disulfide that resulted in complete attenuation of its signal together with formation of reduced glutathione (GSH, m/z 308.091, MH+) and a late migrating reduced thiol analog (m/z 470.144, MH+) not originally detected in DBS extract, and (C) collision-induced dissociation of the precursor ion generating a MS/MS product ion spectrum that is consistent with a mixed oxidized glutathione disulfide containing a modified N-glycated glycine residue.



C. Tyrosine

B. Ophthalmic Acid

A. Serine p =9.16 E-4; FC = 0.76X

p = 7.39 E-4; FC = 0.74X

Relative Ion Response

Relative Ion Response

p = 7.39 E-4; FC = 0.80X Relative Ion Response

CF

SP/non-CF

p = 2.74 E-4; FC = 0.70x

p = 1.13 E-2; FC = 0.84

p = 4.89 E-4; FC =0.75 SN

CF

SP/non-CF

CF

SN

Serine

ROC Curve: CF Vs. SP/non-CF AUC = 0.784 95% CI = 0.694-0.857 p < 0.001

100-Specificity

Ion Response Ratio (Unknown-438)

Tyrosine

SP/non-CF

SN

E. Spearman Rank Correlation Plot

D. Ophthalmic Acid/Unknown-438

Ophthalmic Acid

Sensitivity

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 34 of 36

Presumptive CF by NBS: A: True CF B: Non-CF/Carrier C: Non-CF/IRT

ρ = 0.332 p = 4.55 E-4 n = 108

IRT (ng/mL)

Figure 4. (A-C) Three of the most significant metabolites identified from retrospective DBS cutouts collected shortly after birth that may serve as CF-specific biomarkers for differentiation of CF affected neonates from unaffected carriers, transient neonatal IRT cases and healthy controls when using existing MS/MS infrastructure. In addition to lower concentrations of Tyr and Ser in CF neonates, ophthalmic acid (OPA) was also found to be depleted in CF infants, whereas an unknown trivalent polypeptide (m/z 438.258) was elevated in CF affected neonates. (D) Receiver operating characteristic (ROC) curve highlighting that the measured ratio of ophthalmic acid/unknown peptide provided good predictive accuracy (AUC = 0.784) to differentiate asymptomatic CF neonates from unaffected carriers. (E) The unknown peptide was the only metabolite from DBS extracts that was directly correlated with IRT concentrations suggesting that it may represent a by-product of trypsinogen activation.


Page 35 of 36

Differentiation of Screen+ CF Neonates

Newborn CF Screening: True CF or Unaffected Carrier?

CF; Affected (n=36)

DBS Extract Orthogonal T score [1] (12.9%)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Metabolomics of Dried Blood Spot (DBS) Extracts by MSI-CE-MS DBS#3

DBS#1

+ESI 0 QC

3b

3a

1 : 2

2b

2a

1 : 1

1b

1a

MS

2 : 1

DBS#2

TOC Image


Non-CF; Unaffected (n= 72)

Journal of Proteome Research 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47

ACS Paragon Plus Environment

Page 36 of 36

Metabolic Signatures of Cystic Fibrosis Identified in Dried Blood Spots

Recommend Documents