plasmaCOVID-19中血浆蛋白质组学检测方案(液质联用仪)

PASEF enables high-throughput quantitativelarge-scale plasma proteomics and applicationto COVID-19 Plasma and serum are routinely sampled from individuals and therefore serve asreadily available sources for biomarker research and proteomics is the technology ofchoice for this task. Abstract Keywords：timsTOF Pro， PASEFplasma and serumproteomics， clinicalresearch proteomics However， the dynamic range of.protein concentration inn thesesamples poses a challenge forin-depth proteome quantification.PASEF is a powerful technologyfor rapid in-depth and sensitiveproteome quantification that isalso applicable to biofluids likeplasma and serum. We combined PASEF with high-throughputgradient settings for the analysisof hundreds of plasma and serumsamples and quantified severalhundred protein groups routinelyin 21- and 47-minute runtimes.The timsTOF Pro coupled withthe Evosep One nanoflow UPLCdelivers quantitation with medianCV less than 20% and quantifiedmore than 70 known biomarkers in a study investigating over 700COVID-19 serum samples Introduction Blood plasma is one of the mostaccessible biofluids for biomarkerdiscovery and is the primarysample of interest in clinical pro-teomics research applications，but comprehensive proteome Figure 1. Plasma proteome workflow. Samples were digested with iST sample preparation kits and purified peptides were measured using a DDA-PASEFmethod. For depletion of abundant proteins plasma samples were loaded sequentially through top-6 and top-14 depletion columns followed by digestionand peptide purification. analysis of plasmaiss challengingdue to the large dynamic range ofprotein concentrations. One commonstrategy to circumvent the dynamicrange problem of plasma proteins isto deplete the most abundant proteinsusing antibodies，facilitating tthedetection of lower abundant proteins.However， these depletion strategiescomewiththe disadvantages ofquantitative variability in the depletionprocess between the samples， andsignificant extra work， time， and costwhen analyzing large sample cohorts.In addition， many high abundanceplasma proteins may serve as carriersfor other proteins that are then alsolost in the process. All of these issuesfavor the direct analysis of undeple-ted samples. Parallel accumulation，serial elution， andfragmentation(PASEF) on the timsTOF Pro hasproven to be highly sensitive and fastfor in-depth or near complete pro-teome quantitation [1，2]. Large scaleplasma proteomics is very attrac-tive for biomarker discovery and wehave previously demonstrated thefeasibility and benefits of the rectan-gular strategy for biomarker studiesas opposed to the common triangularstrategyemployed in biomarkerdiscovery and verification [3]. A criticalpre-requisite for such a rectangularclinical proteomics strategy is a robustproteomics platform with sustainedperformance and sensitivity forseveral hundreds if not thousands ofsamples. One of the hallmarks of thetimsTOF Pro is indeed robustness.which makes it highly suitable for largescale proteomics studies on clinical samples. Here we investigate theperformance of the PASEF technologytorhigh-throughputplasma pro-teomics in two gradient lengths anddemonstrate its power for the analysisof large cohorts. Materials and Methods Pooled plasma was purchased fromSigma (Sigma Aldrich， USA) and indi-vidual plasma was extracted froman apparentlyhealthyyindividual.Plasma proteins were digested usingthe iST sample digestion kit fromPreOmics (PreOmics， Germany). Forconstruction of the libraries， plasmaproteins were first depleted through atop-6 depletion column from Agilent(Human-6)) aand the flow throughwas further depletedl throughatop-14 depletion column (Human-14)(Figure 1). The flow through was thenconcentrated in a 3 kDa molecularweight cut-off membrane filter. Thedepleted plasma was then digestedusing the PreOmics kit using trypsin.The purified peptides were then frac-tionated by Strong Cation Exchange(SCX)on aStagelip) intoitthreefractions or Strong Anion Exchange(SAX) into 6 fractions as publishedelsewhere [4]. Separately， the deple-ted protein mix was also separated tosix fractions based on size-exclusionanddl subsequently digested. Togenerate the libraries， the resultingpeptides from each fraction were runon 17.8-minute gradient (21-minuterun time) using the Bruker nanoEluteHPLC system with a Bruker TENcolumn (BrukerrDaltonik GmbH Germany) coupled via a captive spraysource to atimsTOF Pro mass spectro-meter. To generate libraries for slightlylonger gradients， the peptides fromeach fraction were measured withan Aurora 25 cm column (lonopticksAustralia) with a 30-minute gradient(total run time of 47 minutes). Dataacquisition was performed with theDDA-PASEF method and raw datawere processed together with the cor-responding library in MaxQuant [3]. Plasma samplesfitromCOVID-19patients were obtained from thehospital of lLMU Munich. Controlsamples were obtained from PCRnegative subjects. Samples wereprocessed for subsequent proteomicsin a high-throughput manner usingarobotic platformn asdescribedelsewhere [3]..Purifiedpeptideswere loaded on Evotips (EvosepBiosystems， Denmark). An EvosepOne (Evosep Biosystems， Denmark)system was coupled to a timsTOF Prosystem via captive spray ionizationsource. A 8 cm x 150 um columnpackedsiwith1.9 um ReproSil-PurC18-AQ particles (Dr. Maisch) wasused as a separation column with the60 samples per day (SPD) method(21-minute gradients). All raw data were processed usingMaxQuant version 1.6.17.0 and datawere filtered at 1% FDR at PSM andprotein groups level. The MaxLFQalgorithm was used and the matchbetween runs option was enabledto increase identification from thelibrary runs. Figure 3. Protein and peptide identification numbers in libraries obtained with different gradientlengths. These libraries have been uploaded and are available for use by the community. Table 1. Summary of protein quantification depth Gradient Runtime Experiment Protein groups- median Peptides-median 17.8 21 default 373 2490 17.8 21 mob-stringent 327 2397 17.8 21 RT-stringent 311 2303.5 17.8 21 mob&RT stringent 297 2278.5 30 47 default 524 3313.5 30 47 mob-stringent 462 3114.5 30 47 RT-stringent 408 2881 30 47 mob&RTstringent 375 2806 Results and Discussion Fast single-shot analysis of peptidesfrom plasma samples would be thesimplest and most efficient way toperform large-cohort bottom-upproteomic analyses. First we testedthe performance with a Bruker-TENcolumn with a 17.8 minute gradientfor a 21-minute total run time. For theshort gradient we could quantify onaverage about 230 protein groups perrun without any matching betweenruns option and this increased to about250 protein groups after matchingamong the runs without any library.(Figure 2A， Table 1). To improvethe depth of proteome quantitationwe included a library created fromtop14 depleted peptides. The librarywas measured from six fractionsusing the same configuration (withBruker TEN column and matchinggradient) and this resulted in 724 pro-tein groups and 4551 unique peptidesequences. It is important to notethat this library size in terms ofamount of raw data is very compact，however it comes with enoughinformation that it is sufficient toenable a deeper coverage of theproteome. When using this libraryfor matching， the number of proteinsquantified per run on average was well above 350 protein groups with2512 unique peptides. This indicatesthat we could quantify in the range of350 protein groups in very short gra-dients using a library based matchingapproach. Another landmark study onlarge-scale collisional cross-sectionCCS values measured on aaTIMSdevice has shown that CCS valuesacross different measurement condi-tions are highly robust in comparisonto the variation of peptide elutionbehavior [5]. Intrigued by this report，we also investigated the gain bymatching when usingsstringentparameters for either or both mobilityand retention times. For retentiontime the tolerance was reduced from42 seconds to six seconds and formobility the tolerance was reducedfrom 0.05 to 0.01 for stringent condi-tions. The double stringent condition(for both retention time and mobil-ity) resulted in the least number ofLFQ values. In general， reducing theretention time tolerance resulted inhigher loss of quantification values incomparison to the mobility values. Thiscould be explained by the observationthat the CCS values measured on thetimsTOF Pro are robust comparedto LC retention times as reportedpreviously.1. A previous studvyhasdemonstrated that the proteinsquantified after match betweenthe runs are generally reliable [6]. Therefore， by additional stringency inmatching tolerance， we increase theconfidence in the quantified proteins.For all our analyses we narroweddown the retention time tolerancefrom 42 to 6 seconds and thus ournumber of proteins quantified arehighly confident. Next erepeatedthee analyseson a 25 cm Aurora column fromlonOpticks (Australia) with a 30 minutearadient for a total run time of 47 minu-tes. Similar behavior is also observedfor the analysis with the 30-minutegradients. Here the library consistsof six fractions measured in the samecondition as neat plasma peptidesresulting in 6596 unique peptidesand 1070 protein groups. For mobilitystringent conditions on average456 protein groups were quantifiedin each run at an acquisition rate ofabout 30 samples per day (Figure 2B，Table 1). We generated the depletedlibraries for different gradient lengthswith the minimum possible numberof fractions. For example， in additionto the gradients described above wecould achieve close to 1600 proteingroups for a 47-minute gradient (60-minute run time) (Figure 3) if suchalibrary would be needed. These rawdata are available for download andcould be used together with compat-ible DDA runs to improve the depth of quantification in such datasetswithout the necessity to generateand measure depleted peptides forplasma. A larger library for thesegradients would most likely not resultin increased proteome quantificationbut might result in marginal improve-ment of sequence coverage for thequantified proteins. In addition to the depth of coverage，we also observed protein quantifi-cation with low CVs. The medianCV of proteins quantified was 18%(Figure 4A). Given that we employeddata-dependent acquisition (DDA) forthese samples we observed muchhigher data completeness of about91% and the lower abundant proteinstended to have more missing values(Figure 4B). With our optimized librarybased matching strategy we were ableto obtain precise quantitative values forproteins across 5 orders of magnitude.We observed very high correlationacross different runs (median = 0.94)，and ourprecise LC-MSsetupeven allowed the quantification ofsub-clusters of samples that weretreatedwith slightly differentt re-duction alkylation buffers (Figure 4C). These results demonstrate theapplicabilityy oftheettimsTOFFProplatformfor robustandprecisequantification- a critical requirement for large-scale biomarker initiatives.Another key constituent of a reliablequantitative proteomics pipeline is arobust front-end liquid chromatographysystem. The Evosep One system usesa novel way of generating gradientswith a combination of low-pressurepumpsgenerating apre-formedgradient， and subsequent delivery ofthe preformed gradient with a singlehigh-pressure pump， resulting insignificant advances in robustness， inaddition to very low overhead times.The loading of the samples ontoEvotips further enhances the ability toautomate the process， and providesan additional and final cleanup stepprior to injection onto the column，further enhancing robustness. TheEvosep One is thus tailored for clinical research proteomics where robust-ness and throughput are crucial. Theextremely robust pipeline with theEvosep One and the timsTOF Pro hasalso been previously shown to providein-depth reproducible quantification ofproteomes [7]. Very recently we alsoapplied this powerful combinationto investigate proteome alterations inone of the largest serum proteomestudies on COVID-19 infectedpatients which is available as a preprint[8]. We analyzed serum samples from31 patients in up to 54 days of hospi-talization resulting in 458 samples.Samples from PCR negative controlsfrom 262 patients with COVID-19 likesymptoms served as control. Datawere acquireddusingPASEF andprocessed with a matching library [8]. From this set of 720 patient samples，we could quantify 310 ±18 proteinsincluding morethann70) potentialbiomarkers covering five orders ofmagnitude inproteinn abundance(Figure 5). Remarkably no maintenanceor cleaning was necessary during thisstudy which is crucial for obtaininghomogenously high precision proteinquantity values resulting in very lowCVs and the ability to tease out evenminor changes in the protein level. Thisstudy led to the description of highlydetailed longitudinaltrajectorieSof regulatedserum rproteins andpotential novel biomarkers. The abilityto measure similar and larger datasetsusing this pipeline would provideprecious insights about the questionin hand. Conclusion Despite the abundance range of protein concentration of the plasma proteome， the timsTOF Pro platformenables high-throughput and precise quantification of several hundreds of proteins which is pivotal for largescale studies. In measurement times as short as 21 minutes， we could reliably quantify more than 320 proteinsin each sample without depletion， using single-shot injections and label free quantification. Label free quantificationenables undertaking very large-scale studies without complicating planning and having a reference channelwhich is required in multiplexed quantification. Our data further suggests that the proteome could be measuredwith similar depth but in less than half the turnover time per sample because of PASEF further enabling thethroughput of such studies. Future improvements could include dynamic range tackling methods like partialdepletion or enrichment strategies that improve the depth of coverage in plasma. We strongly believe that withsuch further improvements in sample preparation and acquisition strategies in the existing platform， not onlyhundreds but thousands of plasma proteomes could be quantified in a routine manner leading to population-widecohorts tackled with PASEF. Learn More You are looking for further Information?Check out the link or scan the QR code. www.bruker.com/timstofpro References [1] Meier F et al. (2018). Mol.Cell.Proteomics [2] bioRxiv 2020.12.22.423933； doi： https：//doi.org/10.1101/2020.12.22.423933 [3] Geyer et al. (2016) Mol.Sys.Biology [4] Kulak et al. (2014). Nat.Methods [5] Meier et al.(2021).Nat.Communications [6] Lim et al. (2019). J.Proteome.Res [7]B6ruker Application Note LCMS-170 on Evosep+TimsTOF pro for LFQ： www.bruker.com/en/meta/fileadmin.html?q=1878063-Icms-170-dia-pasef-evosep-ebook [8] www.medrxiv.org/content/10.1101/2021.02.22.21252236v1 For more information on services offered by OmicEra， visit their website at www.micEra.com. For Research Use Only. Not for Use in Clinical Diagnostic Procedures. Bruker Daltonik GmbH Bruker Scientific LLC Bremen·GermanyBillerica， MA · USAPhone +49(0)421-2205-0Phone +1 (978)663-3660 However, the dynamic range of protein concentration in these samples poses a challenge for in-depth proteome quantification. PASEF® is a powerful technology for rapid in-depth and sensitive proteome quantification that is also applicable to biofluids like plasma and serum. We combined PASEF with high-throughput gradient settings for the analysis of hundreds of plasma and serum samples and quantified several hundred protein groups routinely in 21- and 47-minute runtimes.The timsTOF Pro coupled with the Evosep One nanoflow UPLC delivers quantitation with median CV less than 20% and quantified more than 70 known biomarkers in a study investigating over 700 COVID-19 serum samples.