Government & Nonprofit

Advances in the application of comprehensive two-dimensional gas chromatography in metabolomics

Due to excellent separation capacity for complex mixtures of chemicals, comprehensive two-dimensional gas chromatography (GC × GC) is being utilized with increasing frequency for metabolomics analyses. This review describes recent advances in GC × GC
of 12
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  Advances in the application of comprehensive two-dimensional gaschromatography in metabolomics Emily A. Higgins Keppler, Carrie L. Jenkins, Trenton J. Davis, Heather D. Bean * School of Life Sciences, Arizona State University, Tempe, AZ, 85287, USA a r t i c l e i n f o  Article history: Available online 24 October 2018 Keywords: Animal modelsBiospecimensBiotransformationComprehensive two-dimensional gaschromatography (GC    GC)Data reporting In vitro  analysesMetabolomicsMultitrophic interactionsSampling a b s t r a c t Due to excellent separation capacity for complex mixtures of chemicals, comprehensive two-dimensional gas chromatography (GC    GC) is being utilized with increasing frequency for metab-olomics analyses. This review describes recent advances in GC    GC method development for metab-olomics, organismal sampling techniques compatible with GC    GC, metabolomic discoveries madeusing GC    GC, and recommendations and best practices for collecting and reporting GC    GC metab-olomics data. ©  2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-NDlicense ( 1. Introduction Metabolism, which is the sum of chemical reactions of an or-ganism, can be investigated at multiple scales, from a singularbiochemical reaction, to metabolic pathways, to cellular, multicel-lular, tissue, organism, and population-scale analyses. As a part of the functional genome [1], metabolic analyses shed light on thetranslationofgenes,transcriptomes,andproteomestophenotypes,and the in fl uence of the environment on this process. Character-izing changes in the metabolome as a function of external or in-ternal perturbations enhances our understanding of howdevelopment, disease, diet, toxins, medications, stress, the micro-biome, etc. govern living systems, and metabolome studies aretherefore relevant to a broad range of the basic biological sciences(Fig.1).Metabolomedataarealsousefulintheappliedsciencesandindustry, and consequently are of high economic importance; forinstance, metabolome data play a central role in the discovery of new pharmaceutical targets and diagnostic biomarkers, in theproduction of fermented foods and beverages, and in the devel-opment of novel biosynthetic pathways and bioremediation stra-tegies (Fig. 1).As de fi ned by Fiehn in a seminal review of metabolic analysespublished in 2002 [2], studies typically fall into four broad cate-gories based on the amount of the metabolome that is character-ized, and the degree to which metabolites are identi fi ed:1.  Target analysis  to measure the substrate or product of anenzyme or group of enzymes,2.  Metabolite pro fi ling   to identify and/or quantify a class of metabolites (e.g., fatty acid methyl esters (FAMEs)),3.  Metabolic  fi ngerprinting   to rapidly classify samples, whereindividual metabolites are not identi fi ed (e.g., via direct-injection mass spectral methods [3]), and4.  Metabolomics  to comprehensivelyanalyze the metabolome (orlarge fractions thereof), including identifying and quantitatingindividual metabolites.Metabolomics aims to universally detect, characterize, andquantify all metabolites in a biological system [4], but of all of the  ‘ omics approaches (i.e., genomics, transcriptomics, prote-omics), metabolomics is the most analytically challenging. LikemRNA transcripts and proteins, metabolites can be present inhugely disparate concentrations, from single molecules to molefractions, and the absolute and relative concentrations arecontext speci fi c. However, unlike nucleic acids and proteins,made up of combinations of 4 and 22 chemical moieties, *  Corresponding author. E-mail address: (H.D. Bean). Contents lists available at ScienceDirect Trends in Analytical Chemistry journal homepage: © 2018 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ( Trends in Analytical Chemistry 109 (2018) 275 e 286  respectively, the metabolome contains thousands to hundreds of thousands of unique chemical species [5]. No single analyticalplatform can separate and detect all metabolites in a specimen,and to-date, even in the extensively studied human metabolomethat is predicted to contain over 114,000 total metabolites, morethan 80% are yet to be detected [5]. The key to advancing the fi eld of metabolomics is developing the analytical tools todetect, identify, and quantify unknown metabolites, the soft-ware tools to manage and process large quantities of rawmetabolomics data, and the chemometric tools to extract in-formation from the data [2,5].Due to excellent separation capacity for complex mixtures of chemicals, comprehensive two-dimensional gas chromatography(GC    GC) is being utilized with increasing frequency for metab-olite pro fi ling and metabolomics analyses [4]. Typically, whencompared to one dimensional gas chromatography (GC), three-toten-fold more peaks are detectable using GC    GC [4], andtherefore, GC    GC metabolomics is rapidly increasing the meta-bolic catalogs for microbes [6,7], plants [8], animals [9], and humans [10]. Here, we review recent advances in GC  GC methoddevelopment for metabolomics, organismal sampling techniquescompatible with GC  GC, and a selection of GC  GC metabolomicapplications and discoveries that, in our opinion, will push theboundaries of their  fi elds. Additionally, we provide recommen-dations and best practices for collecting and reporting GC    GCmetabolomics data and perspectives on the future directions of GC    GC in metabolomics. An excellent review of GC    GC andmetabolomics was published by Almstetter et al., in 2012 [4], sowe have focused our efforts on reviewing studies that have beenpublished since. 2. Method development for GC £ GC metabolomics There have been signi fi cant efforts and advancements increating robust methods for generalized GC    GC analyses.However, there is still great need for the development of methods specialized for metabolomics, particularly validatedprotocols for specimen preparation, sampling, data collection,and data processing. Other reviews in this Special Issue sum-marize advancements in modulators, stationary phases, massspectrometry, and novel instrumentation, but a few studies thatare speci fi cally focused on methods for metabolomics arereviewed here.  2.1. Sampling and sample preparation Analytical robustness in metabolomics is signi fi cantly impactedbyunevenextractionef  fi cienciesacrosschemicalfamiliesaswellassample inhomogeneity in solid and semi-solid specimens. There-fore, optimized methods for sampling and sample preparation arecritically important. Several recent studies explored the impact of sample preparation on the GC    GC metabolomes of tissues andviscous bio fl uids, and provide recommendations for obtainingmore comprehensive and reproducible metabolomes.The metabolic pro fi ling of organs and tissues has been exploredasastrategytoobtainavarietyofinformationabouthumanhealth;however, the amount of blood in the samples can distort the in-formation gained from these approaches. To explore this, Ly-Verdúetal.used GC  GCwithtime-of- fl ightmassspectrometry(TOFMS)to study the effects that phosphate-buffered saline (PBS) perfusionmay have on the metabolite composition of the liver, and whetheror not perfusion may constitute an essential experimental step inliver pro fi ling [11]. Livers were collected from healthy male mice,and were either perfused with PBS or unperfused prior to har-vesting, homogenization, extraction, derivatization, and GC    GCanalysis.Resultsfollowingmultivariateanalysisrevealedmorethan35 metabolites signi fi cantly differed between the pro fi les of unperfused and perfused livers. The authors observed that theGC  GC metabolomes of perfused livers were slightly less variableand concluded that the presence of blood metabolites can interferewith interpreting liver-speci fi c metabolism in some cases. As aresult, the authors suggest that the choice to perfuse organ andtissue samples must be carefully considered in the context of eachstudy hypothesis since the metabolome will be in fl uenced by thepresence or absence of blood.Sputumisanoft-usedspecimenforlungmetabolomicsanalyses,but its high viscosity and inhomogeneity can impact the recoveryand reproducibility of metabolites. To determine the best pre-treatment method for sputum prior to chloroform/methanol/waterextraction, derivatization, and GC    GC analyses, Schoeman, duPreez,andLootscomparedfourprotocolsusingsputumspikedwith Mycobacterium tuberculosis ,thecausativeagentoftuberculosis(TB)[12].Thefourmethodsevaluatedincludedincubationswith1)a1:1v/v ratio of sputum and SPUTOLYSIN ® (a concentrate of dithio-threitol inphosphate buffer), 2) a 1:1 v/v ratio of sputum and 0.5 NNaOHwith20%w/v N  -acetyl- L  -cysteine,3)a1:2v/vratioofsputumand1NNaOH,and4)asimplehomogenizationofsputumwith45%ethanol in a 1:2 v/v ratio. In the  fi rst three methods the pretreatedmixtureswerecentrifugedandthecellpelletsharvestedforfurtherextraction, and in the fourth method the entire homogenate wasretained and dried prior to CHCl 3 /CH 3 OH/H 2 O (1:3:1) extractionand silylation. By analyzing the extraction ef  fi ciency (i.e., numberand intensity of compounds extracted), repeatability, limits of detection(LOD),andthepredictiveaccuracyofbiomarkersselectedfrom the GC    GC metabolomes, they determined ethanol ho-mogenization is the superior pretreatment method, which allowedthem to identify 19 metabolic biomarkers of TB using only 250  m L sputum. While it is not surprising that the ethanol extractionmethod, which retains the entire sputum sample, produces thegreatest number and concentration of metabolites (roughly 80%more than SPUTOLYSIN ® , in second place), it is interesting to notethat it also generated metabolite pro fi les that were highlydiscriminatory between the  M. tuberculosis  spiked sputum vs.unspiked controls and these metabolites correctly classi fi ed TB-positive vs. TB-negative patient samples. These data suggest thatthesecreted M. tuberculosis metabolitesserveas usefulbiomarkers,not just the intracellular metabolites, which may be the key todiagnosing the presence of TB disease using sputum specimenswith typically low bacterial cell densities ( < 10 5 cells/mL). Fig. 1.  Metabolome data inform the in fl uences of internal and external perturbationson biological systems (dark boxes, left), and have industrial, technological, and medicalapplications (light boxes, right). E.A. Higgins Keppler et al. / Trends in Analytical Chemistry 109 (2018) 275 e  286  276  The success of untargeted metabolomics studies that utilizerelative quantitative data (e.g., multi-marker studies, biomarkerpanel discoveries) relies upon the ability to reproducibly andquantitativelyextractawidevarietyofmetaboliteswhilemitigatingmatrix effects. P  erez Vasquez, Crosnier de bellaistre-Bonose et al.developed a novel double extraction method to capture additionalurine metabolites, and analyzed the derivatized compounds byGC   GC-quadrupole MS (qMS) [13]. The  fi rst urine extraction wasmodi fi ed from a commonly used procedure wherein urea isremoved via urease incubation, then theorganic acids are extractedviaa liquid-liquid extractionwith ethyl acetate. They performed thesecond extraction on the remaining aqueous phase, which was  fi rstincubated with triethylamine at pH 9, then extracted recursivelywith tetrahydrofuran. The organic phases from both steps werepooled and silylated for analysis. This time-intensive sample prep-arationprotocolcombinedwithgreaterpeakresolutionachievedbyGC  GC-qMS facilitated the detection of 92 additional compoundsinurinecomparedto a commonly used sample preparation methodand GC-MS analysis. The validated GC  GC method is used in theirhospital to analyze urine samples of children with neurologicaldisorders of unknown srcin, and the authors posit that theirapproach may be adaptable for metabolic pro fi ling of other body fl uids, such as cerebrospinal  fl uid, saliva, or breath condensates.Marney, Synovec, and colleagues explored how the ratio of sample mass to solvent volume impacts the extraction ef  fi ciency of soluble metabolites from mouse heart tissues by measuring theGC  GC-TOFMSsignalintensityofeightrepresentativemetabolites:fumarate, malate, glutamate, citrate, succinyl-CoA, myo-inosotol,glycerol-3-phosphate, and glycerol [14]. By recursive extractionexperiments on 40 mg tissue in 1 mL solvent (3:1:1 v/v/v CHCl 3 /CH 3 OH/H 2 O), they determined that  fi ve of the metabolites werequantitativelyextracted on the fi rst round, while fumarate, glycerol,and citrate required four to  fi ve extractions to achieve quantitativeresults. To determine a ratio of tissue mass to solvent volume thatwouldyieldmoreuniversallyquantitativeextractioninasinglestep,they measured the linearity and reproducibility of each metabolitewhen using 2 mL solvent to extract tissue at four masses rangingfrom 5 to 40 mg. They found that a 20 mg sample provided anaverage relative standard deviation (RSD) of 20 e 30% in theirmetabolomics analyses, which was suf  fi ciently low to detect rele-vant metabolic changes in their experiments. These data show thateffortstooptimizeextractionef  fi ciencyandenhancereproducibilityprior to specimen analysis will yield higher quality relative quanti-tation data when extraction is a signi fi cant source of variation in anexperiment.Uniform extraction ef  fi ciency is also a concern for volatilemetabolomics analyses in which a sorbent is used for sampling.Solid-phase microextraction (SPME) has become an essential gasphase sampling technique, and several sorbents are commerciallyavailable to optimize sampling for each investigation. Purcaro et al.conducted an analysis of   fi ve different SPME  fi bers consisting of combinations of three sorbents  e  divinylbenzene (DVB), carboxen(CAR), and polydimethylsiloxane (PDMS)  e  to determine the best fi berand sampling conditions foranalyzing the volatile metabolitesof cell cultures infected with human rhinovirus [15]. Based on thenormalized peak areas of 12 volatile and semi-volatile standardsextracted from the cell culture media at 43  C for 30 min, theyidenti fi ed the DVB/CAR/PDMS triphase  fi ber as the best option fortheir analyses as it yielded the highest chromatographic peak areas.They further optimized their sampling method by using a centralcomposite design and response surface modeling to identify thebest time (15 e 45 min) and temperature (37 e 50  C) combination toyield the highest peak intensities for each individual standard.While six of their standards were modeled to produce the highestpeakareaswhensampledat43  Cfor30min,theothersixstandardsdid not produce peak area maxima, and therefore quantitativesamplingmaynotbeachievableundertheexperimentalparametersthat were tested. Because SPME sampling is often performed inparalleltoGC  GCanalysis,therearepracticallimitstothelengthof the sampling period, which are usually limited by the duration of the GC    GC runtime. Therefore, unequal extraction rates are asigni fi cantconcernforrelativequantitationbySPME.Anotherfactorthat can impact quantitation is interanalyte displacement, whichhas been of persistent concern for DVB/CAR/PDMS triphase  fi bers,and could have played a role in the differences in optimal samplingconditions for the 12 standards used by Purcaro and colleagues. Todetermine the extentof this problem among SPME  fi bers, Risticevicand Pawliszyn analyzed the performance of seven commercialphases, measuring analyte extraction ef  fi ciency and sensitivity,desorption carryover, linear dynamic range, and interanalytedisplacement by performing headspace (HS)-SPME on apple ho-mogenatesandanalyzingthevolatilepro fi lesusingGC  GC-TOFMS[16]. The DVB/CAR/PDMS triphase coating outperformed otherphases on extraction ef  fi ciency, as reported by Purcaro and col-leagues[15],butwerealsomorepronetocarryoverandinteranalytedisplacement of a subset of metabolites in the apples. However,decreasing the extraction time signi fi cantly improved both issues,with the tradeoff of increasing the LOD for some analytes.The physical properties of samples (e.g., ionic strength) alsoin fl uence the number and concentration of volatile metabolitesthat are detected, and salt is routinely used to increase the parti-tioning of semi-volatile compounds from the liquid phase to theheadspace. In a comprehensive analysis of urine volatile metab-olomes by HS-SPME and GC    GC-TOFMS, Rocha et al. consideredthe in fl uence of pH on metabolite detection [17]. When comparingchromatograms of aliquots of the same urine sample at pH 5.8(physiological), pH 2.0, and pH 12.0, more than 40% of theapproximately700GC  GCpeaks couldbetentativelyidenti fi edinthe pH 2.0 and pH 12.0 samples, whereas only 163 compoundswere identi fi ed in the physiological sample (pH 5.8). The highestchromatographic area and compound numbers were obtainedunder acidic conditions; therefore Rocha et al. concluded thatuntargetedurine volatile metabolomics should be performed at pH2.0. However, they also noted that targeted analysis or metabolitepro fi ling might be more appropriate at a higher pH, depending onthe metabolites of interest.  2.2. Instrumentation The majority of GC    GC analyses  e  metabolomics analysesincluded  e  are conducted using cryomodulation, which generatespeak widths on the order of 100 ms. The narrowness of the peaksnecessitatesusingMSdetectorsthatcancollectfullscanspectraatarate of 100 Hz to facilitate accurate peak quantitation and decon-volution.TOFMSisthemostcommonmethodofionseparationusedwithGC  GC,capableoffullspectrumcollectionratesupto500Hz,but these instruments are expensive, which limits accessibility.Compared to TOFMS, qMS instruments are comparatively inexpen-sive, generally have a smaller footprint, and provide lower LODs viaselected ion monitoring (SIM). However, typical  “ fast ”  qMS in-struments have maximum acquisition rates of 20,000 amu/s, andtherefore themass spectralscan rangewillbe restricted to 200 amuto meet the minimum scan rate for cryomodulation, which is oftentoo narrow for metabolomics analyses, but can be suf  fi cient formetabolic pro fi ling.GC    GC  fl ow modulation is gaining in popularity and marketshare due to the signi fi cant advantage that it reduces consumablecosts by forgoing the need for cryogens. Flow modulationproducesbroader peaks than cryomodulation, which reduces peak capacityand increases LODs, but the wider peaks are more compatible with E.A. Higgins Keppler et al. / Trends in Analytical Chemistry 109 (2018) 275 e  286   277  qMSdetectors.Tranchida,Mondello,andcolleaguesutilizedqMSina study to optimize a  fl ow-modulated GC    GC method for themetabolic pro fi ling of FAMEs [18]. After optimizing column di-ameters, gas  fl ows, temperature programming, and modulationperiods, they identi fi ed FAMEs in  fi sh oil and human serum withlimits of identi fi cation in the range of 100 e 200 pg on column, andlimits of quanti fi cation (LOQ) as low as 3.4 pg in SIM mode. TheseresultsdemonstratethatGC  GC-qMSiswellsuitedformetabolitepro fi ling, using only a few microliters of bio fl uid or micrograms of cells for analysis. Weinert et al. set out to optimize GC    GCequipped with a fast-scanning qMS detector for large-scale untar-geted metabolomics, and compared their results to TOFMS [19].Their GC  GC-qMS method provided good separation in under anhourfor90%of theurineanalytesdetectedbyTOFMS.Scanningtherange of   m/z   60 e 500 at the maximum rate on the qMS (20,000amu/s),theytypicallyobtained7 e 9datapointsperpeakabove10%peak height, providing good peak area and height precision (2.7%and 2.4% mean RSD, respectively). A potential concern with qMS ismass spectral skewing, which negatively impacts mass spectral li-brary matching and peak alignment across chromatograms. Wei-nert et al. quantitated skewing by a variety of metrics, observing15% mean RSD for apex spectra relative intensities (range6.0 e 29.8%) when they included trace-level peaks, and 10% RSD(range5.9 e 21.6%) when trace peaks wereexcluded.While skewingwas not insigni fi cant, the quality and reproducibility of the apexspectrawassuf  fi cientforaligningthemajorityofpeaksacrosstheirsamples.The application of high-resolution mass spectrometry (HRMS) toGC    GC metabolomics is in its infancy, representing only threepercentofthemetabolomicsstudiespublishedsince2012(AppendixTable) [20 e 22]. The impactof using HRMS is greatest for untargetedmetabolomics, where the accurate mass data provides molecularformulasforunknowncompounds.However,theiruseremainsnichebecauseGC  GC-HRMSinstrumentsareexpensive,precludingthemfrom being purchased by most independent investigators. Further,the high-resolution analyses generate very large data  fi les, whichmakes them less amenable than nominal mass detectors for large-scalecomparativemetabolomicsstudiesforbiomarkeridenti fi cation.AsGC  GC metabolomics studies mature to the point of con fi rmingthechemicalidentitiesofmetabolitesthatwereputativelyidenti fi edin nominal mass analyses, the logical next step will be to obtain ac-curate mass data, and with that the proportion of publications thatinclude GC  GC-HRMS data will increase.Compared to MS detectors, vacuum ultraviolet absorptionspectroscopy (VUVAS) has two signi fi cant hardware advantages: asmall footprint and a lack of intensive vacuum requirements.Gruber, Groeger et al. used a cryomodulated GC    GC-VUVAS toanalyzefourbreathsamplesfromanindividualbeforeandduringaglucose challenge [23]. Results showed that detection with VUVAS,with selective monitoring for aromatics, provided similar perfor-mance to GC    GC-TOFMS, and gave good detection for small-oxygenated volatile metabolites (e.g., alcohols and ketones).  2.3. Data processing, analysis, and visualization The ultimate goal of any GC    GC metabolomics analysis is toturn the data collected into chemical and biological information,which is strongly dependent upon reliable methods for processing,analyzing, and visualizing the data. The development of methodsfor GC  GC data processing and analysis is a rapidly-growing area,including novel approaches designed for metabolomics, or vali-dated using metabolomics data [24 e 31]. Because this importanttopic is outside the authors ’  area of expertise, we refer readers tothe GC    GC chemometrics review in this Special Issue and otherreviews [32] for details on recent advancements and recommen-dations in data processing. 3. Applications of GC £ GC in metabolomics Duetothecomplexityofthemetabolomeandtheheterogeneitythat exists within and between organisms, many metabolomicsstudies are begun using reductionist models (e.g., cell culture), andthen may graduate to more complex model systems (e.g., animalmodels), biospecimens (e.g., urine, blood, tissue), and ultimately,living organisms in natural and arti fi cial environments. While the invitro experiments maylackdirecttranslation to organismal-levelmetabolism in native environments, they do provide importantinformation on fundamentals of metabolism, with broad accessi-bilityandlowcosts(Fig.2).Inthissectionwehighlightapplicationsof GC    GC to  in vitro  cultures, analysis of biospecimens, andorganisms, and we review studies that used interesting biologicaland analytical designs to investigate the underlyingmechanisms of metabolism and the roles metabolites play in multitrophic in-teractions. The handful of studies we review in this section wereselected to demonstrate how GC    GC metabolomics studiescan facilitate discoveries and push the boundaries of their  fi elds.A more comprehensive list of GC    GC metabolomics studiespublished between the end of 2011 and June 2018 is available inAppendix Table.  3.1.  In vitro  analyses The recent implementation of GC    GC for untargeted metab-olomics of bacterial cultures has vastly expanded the volatilemetabolome (or  “ volatilome ” ) catalog forhumanpathogens, whichhave been studied for decades using GC-MS. Bean, Dimandja, andHill pioneered the use of GC  GC for untargeted bacterial volatilemetabolomics with a characterization of the volatilome of   Pseu-domonas aeruginosa  strain PA14, detecting 56 chromatographicpeaks associated with the bacterium, which nearly doubled thepublishedvolatilomeof thiswell-studiedorganism[33].Theabilityto detect more chemical diversity in  in vitro  samples via GC    GC Fig. 2.  There are trade-offs between feasibility (i.e., costs, sample access) and trans-latability to living organisms for metabolomics experiments conducted with  in vitro cultures, animal models, biospecimens collected non-invasively (e.g., urine, breath) orinvasively (e.g., tissue biopsy), and  in vivo  or human studies. E.A. Higgins Keppler et al. / Trends in Analytical Chemistry 109 (2018) 275 e  286  278  has facilitated the exploration of the biological diversity withinspecies, and underscores the degree towhich study design impactsthe volatilome. In order to investigate strain-to-strain diversity,Bean, Rees, and Hill used GC  GC tocomparethe volatilomes of 24clinical isolates of   P. aeruginosa  [7]. They were able to detect 391chromatographicpeaksassociatedwith P. aeruginosa ,ofwhichonly70 volatiles were detected in all 24 isolates, termed the core vola-tilome. Using accumulation and rarefaction curves of the pan-volatilome and core volatilome, respectively, they showed thatthey analyzed a suf  fi cient number of samples to capture the vola-tilome diversity of   P. aeruginosa  clinical isolates under the studiedconditions. Their curves also show that to approximate the coremetabolome (with a 50% in fl ation in its size), a median of 12 andminimum of three isolates were required, and the pan-volatilome e  or the collection of all volatile metabolites produced  e  requireda median of 14 and minimum of six isolates to cover 95% of themetabolome. These data demonstrate that de fi ning the metab-olomeof aspecies basedon asingle specimen(orasmall collectionof specimens) is likely to be misleading.Growth conditions can also signi fi cantly in fl uence the microbialmetabolome in in vitro analyses.In astudyof nineclinical  Klebsiella pneumoniae isolatesgrowninfourrichmedia(lysogenybroth,brainheart infusion, Mueller-Hinton broth, and tryptic soy broth), a totalof 365  K. pneumoniae -associated volatiles were detected byGC  GC-TOFMS,ofwhichonly10%wereconservedacrossallmedia[34]. Using principle components analysis (PCA) of the volatilomes,Rees, Hill, and colleagues showed that the bacterial samples clus-tered based on their growth medium and not bacterial strain. This fi nding was true even when only the 36 volatiles that wereconserved across all four media were used as variables in the PCA.Therefore,thevolatilomeof  K.pneumoniae isstronglydependentonthegrowth mediumused,andtheauthorsconclude that thechoiceof medium should be carefully considered in microbial metab-olomics studies. Together Rees's [34] and Bean's [7]  fi ndings un-derscore the challenge in capturing the essence of an organism'smetabolomewithasinglesetofexperiments,muchlessidentifying in vitro  growth conditions that can robustly mimic the  in vivo infectionenvironment.However,theseexperimentsarestilluseful;the more variations in  in vitro  growth conditions we explore, themore we can understand the broad metabolic capabilities of indi-vidual organisms.  3.2. Animal models of human disease Primates, pigs,mice, andratsare used extensivelyin biomedicalresearch to model human diseases and treatments, and the use of GC   GC to measure metabolic changes in these model systems israpidly expanding. Juul and colleagues have used a primate modeland GC  GC-TOFMS to investigate metabolic changes of the fetal-to-neonatal transition in healthy [35] and diseased animals [36]. To establish the healthy metabolome, six late-preterm  Macacanemestrina  were delivered via hysterotomy, with plasma drawnfromcordbloodandeightadditionalpost-birthtimepointsthrough72 h of age. A total of 100 metabolites were identi fi ed, of which 23exhibited signi fi cant changes in concentration over the 72 h sam-pling period and were categorized by their association withsignalingpathways,glucosemetabolism,carbohydrates,andaminoacids [35]. Beckstrom et al. proposed that these metabolites couldbe used as baseline markers of normal birth transition in futureperinatal metabolomics research. Chun, Juul, and colleagues builtupon that hypothesis by utilizing the  M. nemestrina  primate modelto investigate the plasma metabolome of hypoxic ischemic en-cephalopathy (HIE), a common complication of birth that can leadto early and/or long-term neurodevelopmental consequences,includingcerebralpalsyordeath[36].TheyusedGC  GC-TOFMStoanalyze blood samples from 33 macaques that were exposed to 0,15,or18minof  inutero umbilicalcordocclusiontoinduceHIE.Theytreated a subset of the animals by two methods, hypothermia orhypothermia þ erythropoietin,andobtainedserialbloodsamplesatbaseline, 0.1, 24, 48, and 72 h after hysterotomy. They identi fi edtwelve potential biomarkers of HIE that showed statistically-signi fi cant differences between the diseased and control animalgroups. Bycollecting neurodevelopmental data of the macaques upto nine months of age, they identi fi ed eight metabolites that werecorrelated to early and/or long-term outcomes, and four metabo-lites (citric acid, fumaric acid, lactic acid, and propanoic acid) thatpredicted death or cerebral palsy.Mellors, Hill, and colleagues posited that macaques would alsobe excellent models for identifying breath biomarkers of TB fornovel diagnostics in humans [37]. In a feasibility study, they usedGC    GC-TOFMS to analyze breath from three cynomolgus ma-caques ( M. fascicularis ) and two rhesus macaques ( M. mulatta )before and one to two months after  M. tuberculosis  infection. Usingrandom forest (RF) analysis, they identi fi ed 49 compounds  e  rep-resented strongly (65%) by hydrocarbons  e  that signi fi cantlychanged during the course of infection. They demonstrated thatbreath sampling and analysis is feasible in animal models, and thatbreathmetabolitescanserveasusefulmarkersofinfection.Thefactthat similar animal models are being used in diverse GC    GCmetabolomics studies (e.g., the three macaque studies describedhere [35 e 37]), and that metabolomes are being compared acrossmodel systems (e.g., primate, murine [38], and cell culture [39] models of TB) and with human specimens [12,40 e 42], a morecomprehensive view of the animal models ’  applicability to humandiseases can be built.  3.3. Human biospecimens Because blood, serum, and plasma carry metabolites from allparts of the body and are routinely collected in a clinical setting,they are excellent bio fl uids for metabolomic analyses and theidenti fi cation of biomarkers of disease. Winnike, Zhang et al.compared the utility of GC-TOFMS and GC    GC-TOFMS in meta-bolicbiomarkerquantitationusingpooledserumsamplesfrom109individuals,54ofwhomhaveachronicneurodegenerativedisorder[43]. When comparing metabolomic pro fi les between the healthyand unwell subject groups, 23 compounds detected by GC hadstatistically signi fi cant differences, compared to 34 detected usingGC    GC. Similar advantages for metabolite detection wereobserved by Men  endez-Carre ~ no et al., who developed and vali-dated a method using GC    GC-TOFMS for phytosterol oxidationproducts (POPs) in human plasma [44]. Eleven POPs were spikedinto human plasma samples to validate the detection method. TheLODs andLOQs of GC  GC-TOFMSwere foundtobe approximately10-fold lower compared with GC-MS. In addition to the 11 knownPOPs, GC    GC facilitated the identi fi cation and quantitation of unsaturated brassicasterol and stigmasterol, reported in humanplasma for the  fi rst time.Like blood, urine is a rich source of metabolites from the entirebody, and bears some signi fi cant advantages for biomarkerresearchsince itis plentiful andable tobe collectednon-invasively.Zhang, Brenna, andco-workerspublished a pairof studies inwhichthey used GC  GC-qMS with positive chemical ionization (PCI) todetect complexsteroid mixtures in urine of subjects on therapeuticsteroid treatment [45] and of human athletes [46]. The steroids were extracted from urine, derivatized, and analyzed usingGC  GC-qMS using either electron impact ionization (EI), CH 4  PCI,or NH 3  PCI. Ionization with NH 3  preserved structure-speci fi c ionsand the combination with GC    GC facilitated the identi fi cationof endogenous target steroids at physiological concentrations. E.A. Higgins Keppler et al. / Trends in Analytical Chemistry 109 (2018) 275 e  286   279
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks