Journal of Proteomics 125 (2015) 131–139

Contents lists available at ScienceDirect

Journal of Proteomics

j ourna l homepage: www.e lsev ie r .com/ locate / jp rot

Chromatographic behavior of peptides containing oxidized methionineresidues in proteomic LC–MS experiments: Complex tale of asimple modification

YingW. Lao a,MineGungormusler-Yilmaz b,1, Sabbir Shuvo c, Tobin Verbeke c,2, Vic Spicer e, Oleg V. Krokhin d,e,⁎a Department of Chemistry, University of Manitoba, Winnipeg R3T 2N2, Canadab Department of Biosystems Engineering, University of Manitoba, Winnipeg R3T 2N2, Canadac Department of Microbiology, University of Manitoba, Winnipeg R3T 2N2, Canadad Manitoba Centre for Proteomics and Systems Biology, Winnipeg R3E 3P4, Canadae Department of Internal Medicine, University of Manitoba, Winnipeg R3E 3P4, Canada

Abbreviations: SRM, selected reaction monitoring;Acquisition of all Theoretical Mass Spectra; HI, hySequence-Specific Retention Calculator; Mso, methionineMsn, methionine S,S-dioxide, methionine sulfone.⁎ Corresponding author at: Manitoba Centre for Pro

799 JBRC, 715 McDermot Avenue, Winnipeg, R3E 3P4,E-mail address: krokhino@cc.umanitoba.ca (O.V. Krok

1 Present affiliation: Department of Bioengineering, E35100, Turkey.

2 Present affiliation: Biosciences Division, Oak RidRidge, TN 37831, USA.

http://dx.doi.org/10.1016/j.jprot.2015.05.0181874-3919/© 2015 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 22 January 2015Received in revised form 6 May 2015Accepted 20 May 2015Available online 27 May 2015

Keywords:Post-translational modificationsMethionine oxidationPeptide retention predictionReversed-phase HPLCAmphipathic helicity

On average, the oxidation of a single Met residue to Mso (methionine S-oxide, methionine sulfoxide) and Msn(methionine S,S-dioxide, methionine sulfone) decreases peptide retention in RP HPLC by 2.37 and 1.95 Hydro-phobicity Index units (% acetonitrile), respectively. At the same time, the magnitude of the retention shift variesgreatly (−9.1 to +0.4% acetonitrile for Mso) depending on peptide sequence. The latter effects are mostlyassociated with the stabilization of secondary structures upon peptide interaction with the hydrophobic station-ary phase: when an oxidized residue is located in the hydrophobic face of an amphipathic helix, the decrease inretention is profound. The sameamino acid positioning leads to complete or partial resolution of pairs of peptidescontaining diastereomeric Mso residues. Contrary to all previously reported observations, and the nature of thismodification, we also demonstrate for the first time that methionine oxidation may increase peptide hydropho-bicity. This behavior is characteristic for Met residues in the N3 position of an N-capping box stabilization motifprior to the amphipathic helix. All these findings indicate that the prediction of peptide secondary structuresupon interaction with hydrophobic surfaces must become an integral part of peptide retention modeling inproteomic applications going forward.

© 2015 Elsevier B.V. All rights reserved.

1. Introduction

Bottom-up proteomics is the leading tool in modern analyticalchemistry of proteins [1]. Vast majority of such analyses include proteindigestion, reversed-phase LC separation of the resulting peptidemixtures, and MS/MS analysis. The major obstacle encountered inthese protocols is an inadequate depth of analysis due to a wide dy-namic range of protein concentrations [2] and presence of post-translationally/chemically modified peptides. Modified peptides fromthe most abundant proteins overpopulate the analysis space in both

SWATH, Sequential Windowdrophobicity index; SSRCalc,S-oxide, methionine sulfoxide;

teomics and Systems BiologyCanada.hin).ge University, Bornova-Izmir,

ge National Laboratory, Oak

LC and MS dimensions, leading to failure of a mass spectrometer to se-lect the low-abundance peptides forMS/MS acquisition. Chemical insta-bility of the analytes also complicates targeted (selected reactionmonitoring, SRM) and data independent (SWATH) quantitative tech-niques. Peptideswith potentialmodification sites are typically excludedfrom SRM assays [3]. Assignment of modified peptides in SWATHdatasets is often influenced by their modified counterparts, as they usu-ally are detected within the same SWATH window [4]. On the otherhand, monitoring in-vivo PTMs is an important part of modern systemsbiology studies. There is ample evidence that PTMs control the functionsof biological systems [5,6]. All these issues make studying the behaviorof modified peptides in LC–MS analysis extremely important. Knowingthe exact mass shifts induced by these modifications, and the behaviorof the affected peptides upon MS/MS fragmentation is critical to suc-cessful analysis. The application of a second identification constraintsuch as predicted retention time would solidify all MS-based assign-ments and simplify the development of quantitative targeted assays.

Presently, about 500 modifications are known and reported [7].Methionine oxidation is one of the most frequently observed post-translational modifications of peptide/proteins, and may constituteup to 10% of the total non-redundant peptide sequences [8] identified in

H3

O

N

H

C

S

C

O

O

H3

O

N

H

C

S

C

O

H3

O

N

H

C

S

C

OO O*O

*stereogenic center

methionine

+15.995 Da methionine S-oxidemethionine sulfoxide (Mso)

+31.990 Damethionine S,S-dioxide methionine sulfone (Msn)

Fig. 1. Methionine oxidation to methionine S-oxide (Mso) and methionine S,S-dioxide(Msn). Stereogenic sulfur atom is labeled with asterisk.

132 Y.W. Lao et al. / Journal of Proteomics 125 (2015) 131–139

typical bottom-up LC–MS experiments. Methionine S-oxide (methioninesulfoxide, Mso, +15.995 Da, Fig. 1) is themajor product of oxidation.Intense oxidation can cause methionine sulfoxide to react further,producing a methionine S,S-dioxide derivative (methionine sulfone,Msn, +31.99 Da), however this reaction proceeds at a much slowerrate. Many age-related diseases are associated with the accumula-tion of Mso, among them Alzheimer's, Parkinson's, emphysema,bronchitis, rheumatoid arthritis, inflammatory disorders, and cata-ract formation [9]. Unintentional oxidation occurs primarily duringsample preparation and handling, although it has also been reportedthat electrospray ionization may induce oxidation [10]. Usually in-source oxidation can be easily detected based on the identical elutionprofiles of modified and non-modified peptides [11]. Schöneich et al.[12] proposed several possible oxidation pathways that can beencountered during the oxidation of peptides, such as oxidation byfree hydroxyl radicals and metal-catalyzed oxidation with or withoutneighboring group effect. Most interestingly, however, the formationofMso leads to amixture of RS and SS diastereomers, differing in config-uration at the sulfur atom (Fig. 1). Peptides carryingMso diastereomerscan be separated using conventional (achiral) RP HPLC methods onlywhen effects impacting the secondary structure are involved. Blondelleet al. [13] reported that a doublet peakwas observed during the separa-tion of amphipathic helical peptides containing methionine sulfoxidethat was located in the hydrophobic face of the helix. The same posi-tioning caused a significantly larger negative retention time shift uponoxidation. Peak splitting for Mso-containing peptides is routinelyobserved in proteomics experiments [14]. Recently, Khor et al. [15]demonstrated the successful identification and differentiation of theRS and SS diastereomers using methionine sulfoxide reductase en-zymes MsrA and MsrB, respectively.

On the other hand, the use of modified peptides that exhibit differ-ent affinities for the stationary phase in RP chromatography has openedup new avenues of exploration for peptide-centric proteomics [16]. Forexample, Met oxidation has been used in the combined fractionaldiagonal chromatography (COFRADIC) [17,18] technique for theselective isolation of Met-containing peptides. This method utilizes a2-dimensional RP-LC separation under identical conditions, withhydrogen peroxide oxidation applied to fractions collected in the firstdimension. Knowing that methionine oxidation reduces retentionallowed the prediction of the expected elution window of the modifiedpeptides in the second dimension. Given the relatively low abundanceof Met in peptide sequences, this allowed to achieve a greater dynamicrange for the analysis of complex peptidemixtures. Another applicationof this method demonstrated selective isolation of Mso-containingpeptides [19]. Selective reduction using MsrA and MsrB reductaseswas used in this instance, to ensure the complete conversion of bothversions of Mso containing peptides into their Met counterparts. Inthis case, fraction collection in the second dimension of the COFRADICprocedure targeted species exhibiting increased retention.

Systematic large-scale observations on chromatographic behavior ofPTM peptides including Met oxidation appeared following rapiddevelopments in LC–MSproteomics applications [8,10,20–22]. Literatureindicates that peptides containing oxidizedmethionine are more hydro-philic under RP LC conditions than their unmodified analogs. Zybailovet al. [8] summarized the effects of 15 different PTMs on peptide elutiontimes and found that retention increases in the order Mso b Met.Lengqvist et al. [10] investigated the effect of common modificationson retention and isoelectric point values and reached the same conclu-sion. Recently Moruz et al. [22] introduced application of their peptideretention model Elude to 5 different PTMs. The authors collected reten-tion information from proteomic analysis of 12 different organisms andsubsequently modified the peptide retention prediction model to incor-porate PTMs. Some of these runs were performed using the COFRADICapproach, thus the authors encountered oxidation on both Mso andMsn. Model optimization was performed using ~1500 peptides forboth Mso and Msn-containing peptides and showed increase in reten-tion coefficients as follows: Mso b Msn b Met. Despite ample evidenceof significant variation in retention shift upon oxidation provided byBlondelle et al. 20 years ago [13], these effectswere never explored in de-tail. The assumption that the shift in peptide retention is independent ofthemodified residue's position or secondary structure effects is sufficientfor additive prediction models. The accuracy of such models reaches itspotential maxima at 0.93–0.94 R2-value in prediction accuracy. Movingbeyond these values was one of the major motivations for this study.

The long-term goal of our research group is to construct an enhancedpeptide retention prediction algorithm, which will encompass variouschemical and biologically relevant post-translational modifications.Following development of our core Sequence Specific Retention Calcula-tor (SSRCalc) model [23] we began systematic studies of modified pep-tides with N-terminal cyclization at Gln and Caboxamidomethyl-Cys[21], and with 5 different Cys-alkylating agents [24]. These studieshighlighted the importance of amphipathic helicity in correct predictionof peptide retention times. Alpha-helices are one of the major elementsof protein and peptide secondary structure. Yet this feature remainsunincorporated in current peptide retention prediction models. Thehydrophobic character of methionine, its abundance inside amphipathichelices, significant differentiation in retention shifts upon oxidationdepending on secondary structure [13] make this modification a key toprobing the effects of amphipathic helicity on RP LC peptide retention.

The goal of this study was to expand application of our SequenceSpecific Retention Calculatormodel to process peptideswithmethionineoxidation. Priority was given to the ubiquitousMet oxide (Mso) contain-ing peptides routinely observed in LC–MS experiments. We attemptedanalysis of the retention shifts for several thousands pairs of Met–Msocontaining tryptic peptides observed in three 2D LC–MS runs. Retentionbehavior of Msn-containing peptides was investigated on a smaller scaleby the 1D LC–MS analysis of tryptic digests oxidized with H2O2. Our ob-servations were also supported by the analysis of designed syntheticpeptides. The overarching objectives of this study included using Metoxidation as a probing tool to better understand effects of helicity onpeptide interaction with hydrophobic surfaces — one of the key mecha-nisms determining peptide interaction in biological systems.

2. Materials and methods

2.1. Chemicals

All chemicals were analytical grade and sourced from SigmaChemicals (St-Louis, MO), unless noted otherwise. Synthetic peptideswere purchased from BioSynthesis Inc. (Lewisville, TX) and JPT (Berlin,Germany), in milligram quantities and in Spike Tide format (50 nmoleper peptide), respectively. HPLC-grade acetonitrile and de-ionized waterfrom Fisher Scientific (Toronto, ON)were used for the preparation of elu-ents. Sequencing-grade modified trypsin (Promega, Madison, WI) wasused for digestion.

133Y.W. Lao et al. / Journal of Proteomics 125 (2015) 131–139

2.2. Protein digestion and sample preparation

Tryptic digests of whole cell lysates of Neurospora crassa,Thermoanaerobacter WC1, and Clostridium butyricum were preparedusing the FASP protocol [25]. Digests were acidified with trifluoroaceticacid and purified by RP solid-phase extraction. Approximately 100 μg ofeach digest (according to NanoDrop 2000 (ThermoFisher Scientific)measurements) were used for 2D LC–MS analysis. Samples of syntheticpeptides were prepared by dilution in Buffer A (0.1% formic acid inwater) to provide injected amounts of ~200–400 fmole per injection.To produce samples with a higher degree of oxidation, and to generateMsn-containing species, the samples were subjected to oxidation with1%H2O2. Depending on the desired oxidation state and degree of oxida-tion in original samples, we varied the hydrogen peroxide exposuretime. The methionine-containing Spike Tide peptides from JPT usuallyexhibit a ~50–60% oxidation degree in Mso without any treatment.Synthetic peptides prepared using the conventional approach, and atryptic digest of Thermoanaerobacter WC1 were each oxidized for upto 6 days to produce sufficient amounts of Msn containing species. Allindividual samples and fractions collected in the first dimension of2D LC procedure were spiked with a mixture of 6 standard peptidesP1–P6 [26] for retention time alignment purposes (400–600 fmole perinjection each).

2.3. Chromatographic conditions

The first dimension separation was performed using RP-HPLC atpH 10, using a method first described by Gilar et al. [27] and modifiedas described elsewhere [28]. Digest aliquots containing ~100 μg ofpeptide mixtures were fractionated. Forty 1-minute fractions were col-lected (roughly 2.5 μg per fraction); these fractionswere then pair-wiseconcatenated to ensure optimal coverage of separation space in the sec-ond separation dimension. One third or each fraction (~1.5–2 μg) wassubjected to 1 h second dimension LC–MS analysis.

Both 1D and 2D analysis procedures employed identical nano-flowRP-HPLC system settings on the final stage of separation prior to MSacquisition. A splitless nano-flow 2D-Ultra system (Eksigent/ABSciex,Dublin, CA) with 10 μL sample injection via a 300 μm × 5 mmPepMap100 (Thermo Scientific) trap-column and a 100 μm × 200 mmanalytical column packed with 5 μm Luna C18(2) (Phenomenex,Torrance, CA) was used. Both eluents A (water) and B (acetonitrile)contained 0.1% formic acid as the ion-pairing modifier. The lineargradients of 0.6–0.8% acetonitrile per minute (0.5–30% B) were usedfor LC–MS acquisitions in 2D analyses, and for the synthetic peptides.A 0.33% acetonitrile per minute linear gradient (0.5–30% B) was usedfor 1D LC/MS analyses of the digests. All gradient runs were completedwith 5 min at 80% phase B and a 7 min equilibration with startingconditions of 0.5% B. Total LC–MS analysis time was ~1 h per fractionin 2D and ~2 h per sample in 1D runs, respectively.

2.4. Mass spectrometry

A TripleTOF5600 mass spectrometer (ABSciex, Concord, ON) wasused in standard MS/MS data-dependent acquisition mode. Survey MSspectra (250 ms) were collected (m/z 400–1500) and followed byup to 20 MS/MS measurements on the most intense parent ions(300 counts/s threshold, +2–+4 charge state, m/z 100–1500 massrange for MS/MS, 100 ms each). Previously targeted parent ions wereexcluded from repetitive MS/MS acquisition for 12 s (50 mDa masstolerance). Following search parameters for identification with X!Tan-dem were used: ±10 ppm and ±30 mDa mass tolerance for parentand fragment ions, respectively; constant modification of Cys withiodoacetamide; variable modifications +15.995 and +29.99 at Met;expectation value cut-off of Log(e) b −3 and b −1 for non-modifiedand oxidized peptides, respectively.

2.5. Hydrophobicity expression and calculations

Retention times for the standard peptides [14,26] and peptides fromtryptic digests were assigned to each non-redundant identification asthe intensity weighted time average for the two most intense MS/MSspectra. Peptide retention was expressed in Hydrophobicity Indexunits (HI, % acetonitrile) via a least squares linear regression fit usingthe HI values previously assigned to eachmember of the P1–P6mixture[26]. Our SSRCalc algorithmwas used for the HI prediction calculations.Hydrophobic moment values were calculated according Eisenberg et al.[29].

3. Results

3.1. Impact of methionine oxidation on the output of LC–MS proteomicanalyses

Table 1 shows the summary of protein and peptide identificationfor one- and two-dimensional LC–MS analyses of the digests used inthis study, including the detection of peptides containing methio-nine sulfoxide. A coverage of 60–70% for the bacterial proteomes in2 dimensional runs over ~20 h of MS acquisition was found typicalfor our Triple TOF5600 platform. The percentages of Mso-containingpeptides in C. butyricum and Thermoanaerobacter WC1 digests typicalfor 1D and 2D analyses performed in our laboratory were ~3% and~10%, respectively. As a comparison, Zybailov et al. [8] reported ~10%of peptides carrying methionine sulfoxide. Their 1D LC–MS runs,however, were performed for in-gel digested proteins, which usuallyresults in additional methionine oxidation. The higher degree ofoxidation in the N. crassa digest in our instance is likely due to differ-ences in sample preparation. In all cases, the two-dimensional sepa-ration increases dynamic range of MS detection making lower-abundance species accessible; this results in better detection ofMso-containing peptides for all three samples. Additional sourcesof this increase include oxidation of peptides after collecting the frac-tions in the first dimension, as they were pair-wise concatenated andlyophilized.

It should be noted that we did not detectMsn containing peptides inthe samples shown in Table 1. Zybailov et al. [8] found “Dioxidation(M) [+31.99]”modification, which correspond to oxidation of 2methi-onines in the same peptide, rather than the formation of methioninesulfone.

3.2. Peak splitting for Mso-containing diastereomers: confirmation ofliterature data using synthetic peptides

The addition of an oxygen atom (Fig. 1), as Met is oxidized toMso,results in increased polarity of the side chain leading to decreased RPHPLC retention of the modified peptide. According to Blondelle et al.[13] this decrease in retention is more profound when the modifiedresidue is situated in the peptide's preferred binding domain — thehydrophobic face of amphipathic helix. The same positioning leadsto peak splitting for Mso-containing peptides. Addition of the secondoxygen (Fig. 1) eliminates the stereogenic center, which should leadto the elimination of peak splitting. To verify this, we used oxidationwith hydrogen peroxide to prepare mixtures containing sufficientamounts of all three versions (Met, Mso and Msn-containing ana-logs) of the synthetic amphipathic peptides ASGVSSVMSSVVGK andASGVSMVVSSVVGK. Both peptides contain a comparable number ofhydrophobic residues and exhibit well defined hydrophobic faces,as shown by their helical projections in Fig. 2. These two identicalfeatures were introduced to generate sequences with comparablehydrophobicity and to highlight the influence of Met positioningwithin the amphipathic structure. The following observations can

Table 1Detection of methionine oxidation in typical 1D and 2D LC–MS bottom-up proteomic experiments.

Sample # of spectra/# of identifiedpeptidesa

Identifiedproteinsb

# of unique non-modifiedpeptides

# of unique Met-containingpeptides

# of unique Mso-containingpeptides

Oxidationratec

1D LC–MSC. butyricum 34182/20001 912 6565 2204 222 3.4%N. crassa 15398/10712 1153 5979 1419 745 12.5%Thermoana erobacter WC1 35886/20939 1101 6969 1957 224 3.2%

2D LC–MSC. butyricum 301661/164756 2174 25027 8800 2880 11.5%N. crassa 306150/180347 4288 41504 11693 7441 17.9%Thermoana erobacter WC1 393105/202969 s 1998 26633 8567 2733 10.3%

a Expectation value cut-off Log(e) b −3 for non-modified peptides and Log(e) b −1 for Mso-containing peptides.b Expectation value cut-off of Log(e) b −3.c Oxidation rate: ratio of Mso-containing peptides to total number of unique peptides.

134 Y.W. Lao et al. / Journal of Proteomics 125 (2015) 131–139

be made based on the corresponding extracted ion chromatogram(XIC) shown in Fig. 2.

a) Retention time (hydrophobicity of a side chain) increases in thefollowing order: Mso b Msn b Met, which agrees with Moruz et al.[22] observations.

b) A much larger retention time shift is observed when the Mso waspositioned in the hydrophobic face of the amphipathic alpha helicalstructure as compared to that on the hydrophilic face: −5.6 and−1.8% acetonitrile for ASGVSSVMSSVVGK and ASGVSMVVSSVVGK,respectively.

c) Peak splitting only occurs when Mso is positioned in the hydropho-bic face of an amphipathic helix.

d) Further oxidation tomethionine sulfone results in the elimination ofpeak splitting.

Manual inspection of the MS data from the proteomics runs inTable 1 confirms these findings (not shown) regarding the Mso–Metpair: larger negative retention shifts are associated with peak splitting.

15 25

15 25

m

m

ΔHI -5.6%

ΔHI -1.8%

A

C

MetMso

Mso

Msn

Msn

Rel

ativ

e in

tens

ityR

elat

ive

inte

nsity

Fig. 2. Retention time shifts upon oxidation of Met residue to Mso andMsn depending on residface of amphipathic helix (ASGVSSVMSSVVGK) leads to significant decrease in retention, and seface of amphipathic helix (ASGVSMVVSSVVGK). OxidationwithH2O2was used to generate samthese separations was 0.78% acetonitrile per minute. Therefore retention time shift of 7.2 min

Due to the small difference in retention between pairs of peptides con-taining Mso diastereomers, these effects are largely ignored withinlarge-scale proteomics runs. Both peptides have identical MS/MSspectra and are usually reported as a single detectable feature bypeptide identification engines.

To rationalize the observations in Fig. 2, it is important to recognizethat depending on the peptide sequence, the hydrophobic Met couldplay different roles upon interaction with the hydrophobic C18 surface.When placed in a preferred binding domain (the hydrophobic face of anamphipathic helix) it provides additional helix stabilization due to thehydrophobic character of its side chain. Consequently, when Met isoxidized to Mso, the essentially more hydrophilic side chain disruptsthis stabilization (Fig. 2 A,B). Oxidation of a Met residue outside thepreferred binding domain has less effect on the peptide retention time(Fig. 2 C,D). Due to the close “intimate” [13] contact of the preferredbinding domain with the hydrophobic surface, even a small alterationin peptide chemistry in this domain substantially impacts the interac-tion energy. In case of theMsodiastereomeric pair, this results in a chiralseparation on an appropriate chromatographic system.

ASGVSSVMSSVVGK

ASGVSMVVSSVVGK

1

2

3

45

6

7

8

9

10

11

12

1314

M

A

S

G

V

V

V MS

S

S

GS K

V

1

2

3

45

6

7

8

9

10

11

12

1314

A

S

G

V

V

VS

S

S

GK

V

V

in

in

B

DMet

ue's location within amphipathic helical structure. A, B— oxidation of Met in hydrophobicparation of peptides containingMso diastereomers; C, D— oxidation ofMet in hydrophilicple containing comparable amounts of peptides in all oxidation states. Slope of gradient for(A) is equivalent to 5.6% acetonitrile.

Table 2Groups of Met–Mso pairs with similar structural characteristics and their correspondingretention time shifts.

Peptide origin Sequence Retention shiftΔHI = (HIMso–HIMet)(% acetonitrile)

(a) N-terminal Met: small negative retention shifta

Thermoanaerobacter WC1 MAGSHTDITER −0.66N. crassa MLCSDWANIKPDIITLGK −0.60C. butyricum MAQSENPDIIFGTDPDCDR −0.47

(b) Met adjacent to a positively charged residue: small (moderate) negative retentionshift

N. crassa IEEHVMKPSR −0.85C. butyricum ALFIPNGTWLPDEMKDAPR −0.84

(c) Met in hydrophobic face of ampthipathic helix: large negative retention timeshift, associated with separation of diastereomeric pair of Mso-containing peptidesa

Synthetic ASGVSSVMSSVVGKb −5.60Synthetic GSIVEEMEEVIGEGERc −9.05Thermoanaerobacter WC1 SETIMEAVK −3.52Thermoanaerobacter WC1 EGYETLLNTDMELELDNLAR −3.39C. butyricum EAITMVVNDMELLSAVTKd −4.17C. butyricum EIIEEMEAFLK −8.08Thermoanaerobacter WC1 VGENLAEDIEVMEAIFEVTK −6.05C. butyricum LFLEDMESFLK −5.11

(d) Met in the hydrophilic face of the amphipathic helix: small (moderate) negativeretention shift

Synthetic ASGVSMVVSSVVGKb −1.79C. butyricum EAITMVVNDMELLSAVTKd −1.28N. crassa GVDMFIESLAR −0.71C. butyricum SLNEALEEIGMVVEGVK −0.72Thermoanaerobacter WC1 VYIIDEVHMLSTGAFNALLK −0.80

(e) Met in the N3 position of N-capping box prior to amphipathic helix: small positiveretention shift

Synthetic GSIVMEIEEVIGEGERe,f 0.36N. crassa SLEMELESLQGNAVRf 0.01N. crassa TLQMLITELSPDTRf 0.13Thermoanaerobacter WC1 TLEMILADTESEQYKf 0.30

a Shown in Fig. 3c.b Shown in Fig. 2.c Shown in Fig. 4d.d Pair of identical peptideswith different location of oxidizedMet residue (underlined).e Shown in Fig. 4c.f N-cap box motifs are underlined.

135Y.W. Lao et al. / Journal of Proteomics 125 (2015) 131–139

3.3. Impact of methionine oxidation on peptide RP-LC retention: positionaland secondary structure effects

Weused 2D LC–MS identifications from Table 1 to compile retentiondata sets of 6421 pairs of non-modified peptides and their counterpartscontaining one Mso residue (Supplementary Table 1). This allowsfor the uniform comparison of retention shifts, which might revealadditional features required for proper algorithm implementation.

Fig. 3 shows the dependence of retention time shifts expressed in HIunits from peptide molecular weight (a) and hydrophobicity (b). Threetrends are worth noting from these dependencies:

a) Retention shifts caused by Met oxidation range from +0.3 to−8.1% acetonitrile, or −2.37% on average. In most cases, oxidation ofMet to Mso reduces peptide hydrophobicity (retention in RP HPLC).However we found five instances of positive retention shifts.

b) Met oxidation shows smaller effects for large peptides (Fig. 3 A),which is consistent with our previous observation for other modifica-tions [21,24]: the modification of single residues induces smallerchanges in larger peptides.

c) Extremely large retention time shifts (5–8% acetonitrile) werefound only for relatively hydrophobic peptides (HI N 10% ACN inFig. 3 B).

The last observation provides additional insight into the effect ofpeptide amphipathic helicity on peptide RP HPLC behavior. Thepresence of methionine in the hydrophobic face of an amphipathichelix is a prerequisite for a significant retention time shift. Thereforethese peptides must exhibit a well-defined hydrophobic face, e.g. con-tain several hydrophobic residues in their sequences. Because of that,relatively hydrophilic species (with HI b 10) do not fit this criterion asthey simply contain a relatively low number of hydrophobic aminoacids. Fig. 3 a also shows the average values of the Eisenberg's hydro-phobic moment [29] for groups of peptides, depending on their reten-tion shifts (6–7, 5–6, 4–5% ACN, etc.). Clearly, the amphipathicityincreases for peptides with large negative retention shifts.

To further elucidate the positional and structural effects of methio-nine oxidation we analyzed retention shifts for five different groups ofpeptides, examples of which are shown in Table 2. The variation in re-tention time shifts was explained based on our current understandingof separation mechanisms according to the SSRCalc model, and newfindings associated with effects of secondary structure:

a) Small negative retention shifts (or small positive on two occa-sions) are often associated with Met being in an N-terminal positionsuch as MAGSHTDITER from N. crassa (green circles in Fig. 3 C). This ef-fect is caused by the ion-pairing mechanism in peptide RP HPLC. Thepositively charged N-terminal amino group interacts with hydrophilicformate counter ions, which provide a “shielding” effect of neighbor

Peptide HydrophobicityMolecular weight (Da)

Ret

entio

n sh

ift (

% A

CN

)

Ret

entio

n sh

ift (

% A

CN

)

A B

-7<ΔHI<-6 μ = 3.13

μ = 3.02

μ = 2.73μ = 2.29

μ = 2.13

-9

-8

-7

-6

-5

-4

-3

-2

-1

0

11500 2500 3500 4500

-9

-8

-7

-6

-5

-4

-3

-2

-1

0

15 10 15

Fig. 3. Retention time shifts (ΔHI) due to oxidation of single methionine into Mso depending opeptide amphipathicity (average hydrophobicmoment μ) for peptides with large decrease in refor subsets of peptides with specific properties (detailed in Table 2). Green circles— peptides wMet located in hydrophobic face of helix.

amino acids [30]. This leads to a significant reduction of the apparenthydrophobicity of the residues at the peptide N-terminus and as suchin smaller difference betweenMet (hydrophobic) andMso (hydrophilic)containing peptide analogs.

b) Similar ion-pairing nearest-neighbor effects [23] are responsiblefor the relatively lowdecrease in retention times formethionines neigh-boring positively charged residues (Lys, Arg, His).

Peptide Hydrophobicity (HI, % ACN) (HI, % ACN)

Ret

entio

n sh

ift (

% A

CN

)

20 25

-9

-8

-7

-6

-5

-4

-3

-2

-1

0

15 10 15 20 25C

n bulk properties of peptides. A — dependence on peptide size. Note increase of averagetention; B—ΔHI vs. hydrophobicity of non-modified peptides; C—ΔHI vs. hydrophobicityith oxidation at N-terminal methionine residue; red circles— amphipathic peptides with

4525 35

4525 35

4525 35

4525 35

GSMVEEIEEVIGEGER

1

2

3

45

6

7

8

9

10

11

12

1314

I

E

EE

I

A

ΔHI -5.5%

V

GG

G

E

R

E

V I

M

EE

1516

1

2

3

45

6

7

8

9

10

11

12

1314

S

I

E

EE

I

BV

GG

G

R

E

V M

EE

1516

GMIVEEIEEVIGEGER

M

ΔHI -1.6%

1

2

3

45

6

7

8

9

10

11

12

1314

S

I

EE

I

CV

GG

G

R

E

V I

EE

1516

GSIVMEIEEVIGEGER ΔHI +0.4%

1

2

3

45

6

7

8

9

10

11

12

1314

S

I

EE

M

DV

GG

G

R

E

V I

EE

1516

GSIVEEMEEVIGEGER ΔHI -9.1%

min

Met

min

min

min

Rel

ativ

e in

tens

ityR

elat

ive

inte

nsity

Rel

ativ

e in

tens

ityR

elat

ive

inte

nsity

Mso

Met

Met

Met

Mso

Mso

Mso

Fig. 4. Retention time shifts upon oxidation of Met residue to Mso depending on positionwithin anN-capping boxmotif (underlined) prior to amphipathic helix: A—GMIVEEIEEVIGEGER: Met residue in N-cap position, hydrophilic face; B — GSMVEEIEEVIGEGER: at N1position in hydrophobic face of the amphipathic helix; C — GSIVMEIEEVIGEGER: at N3position in N-cap box motif, hydrophilic face; D — GSIVEEMEEVIGEGER: Met residue inhydrophobic face and in middle position within a 2-turns helical stretch.

136 Y.W. Lao et al. / Journal of Proteomics 125 (2015) 131–139

c) Extremely large negative retention shifts are always associatedwith Met residues that are part of the hydrophobic face of an amphi-pathic helix. We confirmed this effect by selecting a group of peptideswith extremely high hydrophobicmoment values (μ N 3),which containMet in the hydrophobic face of the helix (red circles in Fig. 3 C).

d) Relatively small negative retention shifts are observed forMet residues located in the hydrophilic face of the amphipathic helix(similar to Fig. 2 C, D). The example of EAITMVVNDMELLSAVTK(C. butyricum) in Table 2with twoMet additionally highlights the differ-ence between groups c and d. Met (5) is located in the hydrophilic andMet (10) in the hydrophobic face of the amphipathic helix. Retentionshifts upon oxidation of the same parent peptide were found −1.3and−4.2% acetonitrile for these two positions, respectively. This exam-ple indicates that the chromatographic behavior clearly discriminatesbetween peptides with different position of the modified residue.

e) Three out of five peptides exhibiting small positive retentionshifts contain a Met residue in the N3 position of the N-capping boxstabilization motif prior to the amphipathic helix. The number ofinstances found in naturally occurring peptides from our biologicalsamples is very small: 3 cases from 6421 pairs. However, they highlightan extreme importance of the structural features for understandingeffects of peptide's helicity. We further elucidated this unexpected be-havior using designed synthetic sequences (see the following section).

Detailed inspection of oxidized/non-oxidized peptide pairs may re-veal additional features related to the effects of secondary structures.For example the amphipathic QMFQQMQQDMQR carrying three Metresidues was foundwith all three potential locations of theMso residue.All methionines are located in the hydrophobic face of the helix, howev-er exhibiting different shifts upon oxidation: −3.4, −5.0 and −3.7%acetonitrile for positions 2, 6, and 10, respectively. This indicates thatresidues located in the middle of helical stretch (Met 6) are moreinvolved in stabilization of the molecules upon interaction with hydro-phobic surface compared to the terminal ones.

3.4. Positive retention time shift upon oxidation of Met in the N3 position ofN-cap box motif: confirmation using synthetic peptides

Out of five peptides with positive retention shifts, two contain N-terminal Met and three (SLEMELESLQGNAVR, TLQMLITELSPDTR,TLEMILADTESEQYK) are amphipathic and contain the common se-quence motif X-Leu-Z-Met, where X is represented by Ser or Thr and Zby Glu or Gln. Recently we demonstrated that similar to helices in pro-teins, amphipathic helical structures upon interactionwith hydrophobicC18 phase are stabilized by N-cap box motifs [31]. This motif consists offour residues with each position showing different preferences forindividual residues. The first (N-cap, according to Richardson andRichardson's terminology [32]) position is preferable for Asn, Asp, Ser,Thr, and Gly. Second (N1) and third (N2) positions are favoring hydro-phobic residues [31]. Forth (N3) position is predominantly occupied byGlu, Gln and Asp.

We found that N-cap box motifs prior to amphipathic helicalstructure significantly increase chromatographic retention due to helixstabilization. This stabilization occurs due to reciprocal hydrogenbonding interactions of the N-cap residue's side chain with the back-bone C = O group of the N3 residue [33], and vise-versa side chain ofN3 residue with the−NH backbone group of the N-cap amino acid.

Three peptideswith positive retention shifts start with typical N-capresidues (Ser, Thr), followed by hydrophobic Leu (N1) and have Metresidue in position #4 (N3): SLEM, TLQM, and TLEM. We believe thatthe positive retention time shift upon Met oxidation is observed dueto better ability of the Mso residue to form hydrogen bond with theN-cap−NH group compared to a non-modified Met, and thus to stabi-lize the amphipathic helix.

We confirmed this finding by analyzing the retention behavior ofoxidized/non-oxidized pairs for four synthetic peptides carryingpredicted N-cap motif with the Met residue placed in the N-cap

(hydrophilic face), N1 (hydrophobic face), N3 (hydrophilic face) andin the middle of the amphipathic helical stretch (hydrophobic face ofthe helix). The GSIVEEIEEVIGEGER peptide with optimal compositionof N-cap box motif (SIVE) prior to two turns amphipathic sequence(IVEEIEEVI) was chosen as a template. To additionally increase thehelical interactions we used amino acids with small compact aliphaticchains to construct the hydrophobic face of the helix (Ile and Val).Fig. 4 shows the respective extracted ion chromatograms and axialhelical projections for these sequences.

GMIVEEIEEVIGEGER has the Met residue in the N-cap position,placed in the hydrophilic face of the helix (Fig. 4 A). This explains therelatively low (−1.6% acetonitrile) decrease in retention upon oxida-tion to Mso and no peak splitting for Mso diastereomers.

GSMVEEIEEVIGEGER has the Met residue shifted into the nextposition compared to the previous molecule. However it was enoughto move it into the hydrophobic face of the helix (Fig. 4B), causing adiastereomers' peak splitting and a significantly larger decrease inpeptide hydrophobicity upon oxidation (−5.5% acetonitrile).

GSIVMEIEEVIGEGER contains the Met residue in the N3 positionof the N-cap box motif (Fig. 4 C). The hydrophobicity of the oxidizedpeptide is higher (+0.4%), which confirms our findings made basedon the analysis of real tryptic digests in Table 2.

137Y.W. Lao et al. / Journal of Proteomics 125 (2015) 131–139

GSIVEEMEEVIGEGER features the Met residue in the middle of boththe hydrophobic face in the axial helical projection and the linearsequence of the amphipathic stretch (IVEEMEEVI). This position playsthe most important role in the stabilization of the amphipathic helixon the C18 surface. Replacement of Met for more hydrophilic Msodecreases the retention time dramatically (−9.1% acetonitrile) andgives baseline resolution of Mso diastereomers (Fig. 4 D).

These four examples illustrate that knowing the rules of helices' in-teraction with the hydrophobic C18 surface has allowed us to designpeptides exhibiting extreme values of retention shifts, both positiveand negative.

3.5. Retention prediction for peptides carrying oxidized Met residues

Incorporation of new (modified residues) into peptide retentionmodeling should reflect allmajormechanisms contributing into peptideretention. In a first approximation these include hydrophobic, electro-static, ion-pairing interactions and effects related to stabilization ofamphipathic helical structures. The latter by itself is determined bydifferent components, as the helical stabilization depends on hydropho-bic interactions of side chains with the C18 phase, hydrogen bondsbetween backbone C = O and N–H groups, and electrostatic interac-tions between basic and acidic residues in hydrophilic face of thehelix. It should be noted that our algorithm for the correct assignmentof the helical stretch and computation of its contribution into thepeptide retention value is still under development, and does notcurrently incorporate the effect of N-cap stabilization. As describedpreviously [23], our retention prediction model was optimized acrossthe parameter space to maximize the R2 value between predicted andobserved retention values for our training set. In this study the trainingset included both modified and non-modified peptides, with the latterproviding an “anchor” for mapping to HI values, as their predictedvalues remained constant within the optimization process. The follow-ing alterations were made to the SSRCalc model to incorporate methio-nine oxidation (both Mso and Msn):

a) Retention coefficients (Rc) were optimized for both Msn and Msodepending on residue position relative to peptide ends [23]. Theseincluded retention coefficients for N-terminal (RcN1, RcN2, …RcN5),C-terminal (RcC1, RcC2…RcC5) and the internal locations of the resi-dues (Rc) (Supplementary Table 2). The differences in hydrophobic-ity values for N-terminal positions are caused by ion-pairinginteractions involving the positively charged N-terminal aminogroup. As expected (Fig. 2), the retention coefficients for our PTMsincreases in the order Mso b Msn b Met. The Rc values for Mso andMsnwere found to be very close to that of Ala— aweakly hydropho-bic residue. Optimization set for Msn-containing peptides included273 molecules identified in 1D LC–MS analysis of oxidizedThermoanaerobacter WC1 digest (Supplementary Table 1).

b) Due to significantly higher hydrophilicity of Mso and Msn residues,the effect of ion-pairing at adjacent Arg, Lys and His residues in thesequence is not as profound. Msn and Mso were assigned to thegroup of hydrophilic residues for which these corrections do notapply.

c) The helical contribution is computed by summing energies of i; i+ 3and i; i + 4 interactions between hydrophobic residues within theassigned hydrophobic face of amphipathic helices. Interaction ener-gieswere optimized for each pair of interacting residues; this optimi-zation was performed against a training set of ~10,000 amphipathichelical peptides (manuscript under preparation). We did not per-form a helicity model parameter optimization for Mso residues:there are not enough amphipathic molecules in our 6421 peptidestudy set. The number Met–Msn pairs used for the optimization ofthe retention coefficients was even lower, and could result in “over-fitting” of the model. Instead, following from (a) we simply assignedhelical interaction energies for bothMso andMsn to be equivalent to

the Ala values by a substitution on the model's sequence input(Supplementary Table 2). A more sophisticated general approachfor incorporating PTMs into helicity prediction is under development.

Variation of the helical contribution depending on position of theMso residue can be illustrated using example of QMFQQMQQDMQRdiscussed above. Three i; i + 3 and i; i + 4 interactions were assignedto hydrophobic face of this peptide: M - - - M, F - - M and M - - M. Ox-idation of terminal Met (2 and 10) will affect only one interaction. Atthe same time oxidation of Met (6) will affect all three, thus reducinghelical stabilization in larger extent.

Fig. 5 A, B shows SSRCalc HI prediction for 6421 Met containing(a) and single Mso containing (b) peptides. The latter shows worsecorrelation and a −2.48 HI units intercept, which is very close to theaverage contribution of single oxidation of −2.37% acetonitrile. Due toreduced hydrophobicity of Mso, the application of a non-modifiedSSRCalc model yields a poor correlation of ~0.90 when applied tocombined set of modified and unmodified peptides (Fig. 5 C). Ourmod-ifications to the SSRCalc model brought accuracy of retention predictionback to a normal (~0.96) value with identical prediction errors for bothmodified and non-modified species (Fig. 5 D). Note that dependence inFig. 5 D includes Mso and Msn containing peptides as well as peptideswith multiple oxidations in the sequence.

4. Discussion

Better understanding the behavior of oxidized peptides in RP LCbenefits many areas of applied proteomics. Establishing the range ofretention time variation and rules governing them would help inoptimizing Met-COFRADIC protocols, improve the accuracy of SWATHquantitation and confidence of peptide identification in shotgunanalyses. We compared our observations and previously reported data[8,13,17,22] on retention shifts by converting retention values inminutesinto HI units (% acetonitrile) using the provided experimental conditions(gradient slopes % ACN/min). The largest negative retention shift wasreported by Blondelle et al. [13] using the Met substitution in synthetic“perfect” amphipathic helical peptide Ac–LKLLKKLLKKLKKLLKKL–NH2

featuring nine Leu and nine Lys in a 3.6 residue periodicity. Retentiontime shifts were measured between −9.3 and −0.5 min, whichcorresponded to −9.3 and −0.5% ACN for 1% per minute gradient.Gevaert et al. [17] used−7 to−1 min (−4.9 to−0.7% ACN) retentionshift window for fraction collection in COFRADIC experiments. Zybailovet al. [8] reported average retention shift of −3.4% ACN with variationbetween −5.0 and −1.8%. Retention shifts in Moruz et al. [22] datawere estimated between −8.3 and −0.4% ACN. Our results (−9.1and +0.4) confirmed the largest retention shifts values observed byothers, and showed for the first time that oxidation may increase pep-tide hydrophobicity.

We also demonstrate for the first time a number of peptides' compo-sitional and sequence features, which affect retention upon Met oxida-tion. For example, small negative (or even positive) retention shiftcould be the result of an N-terminal location of Met or its N3 positionin anN-capping boxmotif. Differentmechanisms in these two instancesresult in similar chromatographic behavior. It is interesting to note thatboth these cases may represent a complicated case for SWATH analysis[4]. Many peptides from these two categories will generate oxidationproducts with m/z values in the same 25 Da SWATH window, andsimilar retention times and y-ions patterns. Distinguishing betweenthem may require additional efforts.

Amphipathic helicity was found to be the major contributor to di-verse effects of Met oxidation on chromatographic properties. Thiswas previously demonstrated using substitution with Met in a modelsynthetic peptide [13]. Our results expand these observations into realtryptic peptides observed in proteomics experiments. The incorporationof helical correctionswill be amajor improvement in further developingpeptide retention models. Our preliminary attempts for incorporating

y = 0.9776x - 1.2222

R2 = 0.898 y = 0.9757x + 0.049

R2 = 0.96

Obs

erve

d H

I (%

AC

N)

Obs

erve

d H

I (%

AC

N)

Obs

erve

d H

I (%

AC

N)

A B

C D

y = 0.9761x + 0.0525

R2 = 0.958

0

5

10

15

20

25

30

35

0 5 10 15 20 25 30 35

SSRCalc HI (% ACN)

y = 0.9886x - 2.4897

R2 = 0.949

0

5

10

15

20

25

30

35

0 5 10 15 20 25 30 35

SSRCalc HI (% ACN)

0

5

10

15

20

25

30

35

0 5 10 15 20 25 30 35

SSRCalc HI (% ACN)

Obs

erve

d H

I (%

AC

N)

0

5

10

15

20

25

30

35

0 5 10 15 20 25 30 35

Modified SSRCalc HI (% ACN)

Fig. 5. Retention prediction for modified and non-modified peptides using SSRCalc algorithm. A— Retention prediction for set of 6421 non-modified peptides used in this study (Fig. 3);B— retention prediction for set of their 6421 oxidized counterparts carrying singleMso residue; C— retention prediction for combined set A and B; D—modified SSRCalcmodel providesidentical accuracies for modified and non-modified species. This plot includes 273 Msn containing peptides and 336 peptides carrying multiple Mso residues.

138 Y.W. Lao et al. / Journal of Proteomics 125 (2015) 131–139

helicity into the SSRCalc model demonstrate that we can achieve iden-tical prediction accuracy for both modified and non-modified peptidesof ~0.96 R2 correlation.

Understanding separation mechanisms and the effect of molecularstructure on chromatographic behavior of biopolymers has served as acentral theme of the research for many separation scientists. Followingoriginal developments in various protein separation techniques,Regnier noted that most of them are surface-mediated separationmethods [34], therefore the structural features of proteins will deter-mine their chromatographic behavior. The reversed-phase separationmechanismwas described as amulti-point interaction of the hydropho-bic sites on a protein surface and the hydrophobic stationary phase [35],where retention is a function of contact surface areawith elution occur-ring due to a stoichiometric displacement between solute and solventmolecules [36]. Karger and co-workers showed that the hydrophobicinteractions with alkyl-silica surface in RP LC changes the structure ofpolypeptides and may induce protein unfolding [37,38]. Similarly, forshorter peptides, hydrophobic chromatographic packing may inducehelical structures in potentially helical molecules [39]. Therefore onecan envision three different types of peptide binding to a hydrophobicsurface: i) where the conformation of the peptide plays no role; ii) apeptide helical conformation is induced to maximize the contact area;and iii) a peptide already contains an amphipathic helix and bindswith little change to its conformation. Multiple studies in the 1990'sdemonstrated such behavior using designed model peptides [39–41]and led to establishing separate hydrophobicity scales for random coiland helical conformations of peptides upon interactionwith RP sorbents[42]. The arrival of proteomic techniques provides chromatographerswith access to thousands of peptides with secondary structures thatproviding a significant contribution to their RP LC retention [31].These studies will continue the search for a better understanding the

underlying mechanisms, and ultimately, to more accurate predictionof peptide retention properties.

5. Conclusions

We used the outputs of large-scale LC–MS/MS analyses to systemat-ically study of the RP-HPLC behavior of peptides containing oxidizedmethionine residues. In addition to confirming the commonly acceptedphenomenonof lowered hydrophobicity uponMet side chain oxidation,we focused on the retention mechanism and the effects of peptide sec-ondary structure. Oxidation to methionine S-oxide (Mso, +15.995 Da)is themost typical modification resulting in a negative (−2.37%) aceto-nitrile retention shift on average. Further oxidation to methionine S,S-dioxide (Msn, +31.99 Da) increases the retention slightly (−1.95%compared to Met). The former represents themost typical modificationobserved in bottom-up proteomics, and was the major focus ofthis study. The latter is relatively rare, and is typically observed whenoxidation is applied explicitly in sample preparation protocols.

A detailed overview of the chromatographic behavior of peptidescarrying oxidized Met revealed an extremely complex picture inwhich multiple interactions (hydrophobic, electrostatic, hydrogenbond) impact peptide hydrophobicity. The amplitude of retention shiftmay vary dramatically: we found extreme cases of this variation from−9.1 to +0.4% acetonitrile. Stabilization of amphipathic helical struc-tures upon interaction with hydrophobic C18 surface (i.e. secondarystructure effect) is the main reason for such variations. Methionine isa moderately hydrophobic residue, which may be placed in the hydro-phobic face of the amphipathic helix — the preferred binding to C18surface domain for amphipathic helical molecules. Any alteration (Metoxidation in this case) in this domain causes a significant change inpeptide retention. Due to the presence of the stereogenic center in the

139Y.W. Lao et al. / Journal of Proteomics 125 (2015) 131–139

Mso structure, this also leads to partial or complete resolution of peakscorresponding to theMso diastereomers. Moreover, for the fist time wedemonstrate thatmethionine oxidationmay increase peptide retention.This phenomenon is also related to amphipathic helicity. When locatedat the N3 position of an N-cap box motif, the side chain of Mso servesas better partner for hydrogen-bond interaction, which additionallystabilizes the helical structure.

Specific properties of methionine (hydrophobic, prone to oxidation)make it a key in the understanding of retention mechanisms and indeveloping better peptide retention prediction models. Collection ofrepresentative retention datasets has been one of the major problemsin this field. Peptide selection based on retention shift upon oxidationwill direct data collection and subsequent model development forextended sets of amphipathic helical peptides, thus contributing tosolving this fundamental problem in peptide chromatography.

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.jprot.2015.05.018.

Acknowledgments

Thisworkwas supported in part by grants from the Natural Sciencesand Engineering Research Council of Canada (RGPIN 355939-2011,O.V.K.) and Genome Canada (MGCB2 project, O.V.K.). The authorsthank Drs. D. Levin, R. Sparling and D. Court (University of Manitoba)for their assistance in providing the C. butyricum, ThermoanaerobacterWC1, N. crassa cell cultures.

References

[1] M. Mann, N.A. Kulak, N. Nagaraj, J. Cox, The coming age of complete, accurate, andubiquitous proteomes, Mol. Cell 49 (2013) 583–590.

[2] A. Michalski, J. Cox, M. Mann, More than 100,000 detectable peptide species elute insingle shotgun proteomics runs but the majority is inaccessible to data-dependentLC–MS/MS, J. Proteome Res. 10 (2011) 1785–1793.

[3] V. Lange, P. Picotti, B. Domon, R. Aebersold, Selected reaction monitoring for quan-titative proteomics: a tutorial, Mol. Syst. Biol. 4 (2008) 222.

[4] L.C. Gillet, P. Navarro, S. Tate, H. Rost, N. Selevsek, L. Reiter, et al., Targeted dataextraction of the MS/MS spectra generated by data-independent acquisition: anew concept for consistent and accurate proteome analysis, Mol. Cell. Proteomics11 (2012) (O111 016717).

[5] J. Seo, K.J. Lee, Post-translational modifications and their biological functions: prote-omic analysis and systematic approaches, J. Biochem. Mol. Biol. 37 (2004) 35–44.

[6] H. Lin, J. Du, H. Jiang, T.P. Begley, Post-Translational Modifications to RegulateProtein Function, Wiley Encyclopedia of Chemical Biology, John Wiley & Sons, Inc.,2007. 1–31.

[7] E. Ahrne, M. Muller, F. Lisacek, Unrestricted identification of modified proteins usingMS/MS, Proteomics 10 (2010) 671–686.

[8] B. Zybailov, Q. Sun, K.J. van Wijk, Workflow for large scale detection and validationof peptide modifications by RPLC-LTQ-Orbitrap: application to the Arabidopsisthaliana leaf proteome and an online modified peptide library, Anal. Chem. 81(2009) 8015–8024.

[9] J.H. Baxter, C.S. Lai, R.R. Phillips, L. Dowlati, J.J. Chio, S.T. Luebbers, et al., Direct deter-mination of methionine sulfoxide in milk proteins by enzyme hydrolysis/high-performance liquid chromatography, J. Chromatogr. A 1157 (2007) 10–16.

[10] J. Lengqvist, H. Eriksson, M. Gry, K. Uhlen, C. Bjorklund, B. Bjellqvist, et al., Observedpeptide pI and retention time shifts as a result of post-translational modifications inmultidimensional separations using narrow-range IPG–IEF, Amino Acids 40 (2011)697–711.

[11] M. Berg, A. Parbel, H. Pettersen, D. Fenyo, L. Bjorkesten, Detection of artifacts andpeptide modifications in liquid chromatography/mass spectrometry data usingtwo-dimensional signal intensity map data visualization, Rapid Commun. MassSpectrom. 20 (2006) 1558–1562.

[12] C. Schöneich, F. Zhao, J. Yang, L. Miller Brian, Mechanisms ofmethionine oxidation inpeptides, Therapeutic Protein and Peptide Formulation and Delivery, AmericanChemical Society, 1997. 79–89.

[13] S.E. Blondelle, E. Perez-Paya, G. Allicotti, B. Forood, R.A. Houghten, Peptide bindingdomains determined through chemical modification of the side-chain functionalgroups, Biophys. J. 69 (1995) 604–611.

[14] M. Grigoryan, D. Shamshurin, V. Spicer, O.V. Krokhin, Unifying expression scale forpeptide hydrophobicity in proteomic reversed phase high-pressure liquid chroma-tography experiments, Anal. Chem. 85 (2013) 10878–10886.

[15] H.K. Khor, M.E. Jacoby, T.C. Squier, G.C. Chu, D. Chelius, Identification of methioninesulfoxide diastereomers in immunoglobulin gamma antibodies using methioninesulfoxide reductase enzymes, mAbs 2 (2010) 299–308.

[16] M.W. Duncan, R. Aebersold, R.M. Caprioli, The pros and cons of peptide-centric pro-teomics, Nat. Biotechnol. 28 (2010) 659–664.

[17] K. Gevaert, J. Van Damme, M. Goethals, G.R. Thomas, B. Hoorelbeke, H. Demol, et al.,Chromatographic isolation of methionine-containing peptides for gel-free proteomeanalysis: identification of more than 800 Escherichia coli proteins, Mol. Cell.Proteomics 1 (2002) 896–903.

[18] K. Gevaert, J. Pinxteren, H. Demol, K. Hugelier, A. Staes, J. Van Damme, et al., Fourstage liquid chromatographic selection of methionyl peptides for peptide-centricproteome analysis: the proteome of human multipotent adult progenitor cells,J. Proteome Res. 5 (2006) 1415–1428.

[19] B. Ghesquiere, V. Jonckheere, N. Colaert, J. Van Durme, E. Timmerman, M. Goethals,et al., Redox proteomics of protein-bound methionine oxidation, Mol. Cell.Proteomics 10 (2011) (M110 006866).

[20] J. Kim, K. Petritis, Y. Shen, D.G. Camp II, R.J. Moore, R.D. Smith, Phosphopeptideelution times in reversed-phase liquid chromatography, J. Chromatogr. A 1172(2007) 9–18.

[21] J. Reimer, D. Shamshurin, M. Harder, A. Yamchuk, V. Spicer, O.V. Krokhin, Effect ofcyclization of N-terminal glutamine and carbamidomethyl-cysteine (residues) onthe chromatographic behavior of peptides in reversed-phase chromatography,J. Chromatogr. A 1218 (2011) 5101–5107.

[22] L. Moruz, A. Staes, J.M. Foster, M. Hatzou, E. Timmerman, L. Martens, et al., Chro-matographic retention time prediction for posttranslationally modified peptides,Proteomics 12 (2012) 1151–1159.

[23] O.V. Krokhin, Sequence-specific retention calculator. Algorithm for peptide reten-tion prediction in ion-pair RP-HPLC: application to 300- and 100-A pore size C18sorbents, Anal. Chem. 78 (2006) 7785–7795.

[24] X. Jiang, D. Shamshurin, V. Spicer, O.V. Krokhin, The effect of various S-alkylatingagents on the chromatographic behavior of cysteine-containing peptides inreversed-phase chromatography, J. Chromatogr. B Anal. Technol. Biomed. Life Sci.915–916 (2013) 57–63.

[25] J.R. Wisniewski, A. Zougman, N. Nagaraj, M. Mann, Universal sample preparationmethod for proteome analysis, Nat. Methods 6 (2009) 359–362.

[26] O.V. Krokhin, V. Spicer, Peptide retention standards and hydrophobicity indexes inreversed-phase high-performance liquid chromatography of peptides, Anal. Chem.81 (2009) 9522–9530.

[27] M. Gilar, P. Olivova, A.E. Daly, J.C. Gebler, Orthogonality of separation in two-dimensional liquid chromatography, Anal. Chem. 77 (2005) 6426–6434.

[28] R.C. Dwivedi, V. Spicer, M. Harder, M. Antonovici, W. Ens, K.G. Standing, et al., Prac-tical implementation of 2D HPLC schemewith accurate peptide retention predictionin both dimensions for high-throughput bottom-up proteomics, Anal. Chem. 80(2008) 7036–7042.

[29] D. Eisenberg, R.M. Weiss, T.C. Terwilliger, The helical hydrophobic moment: ameasure of the amphiphilicity of a helix, Nature 299 (1982) 371–374.

[30] O.V. Krokhin, R. Craig, V. Spicer, W. Ens, K.G. Standing, R.C. Beavis, et al., Animproved model for prediction of retention times of tryptic peptides in ion pairreversed-phase HPLC: its application to protein peptide mapping by off-lineHPLC–MALDI MS, Mol. Cell. Proteomics 3 (2004) 908–919.

[31] V. Spicer, Y.W. Lao, D. Shamshurin, P. Ezzati, J.A. Wilkins, O.V. Krokhin, N-cappingmotifs promote interaction of amphipathic helical peptides with hydrophobicsurfaces and drastically alter hydrophobicity values of individual amino acids,Anal. Chem. 86 (2014) 11498–11502.

[32] J.S. Richardson, D.C. Richardson, Amino acid preferences for specific locations at theends of alpha helices, Science 240 (1988) 1648–1652.

[33] R. Aurora, G.D. Rose, Helix capping, Protein Sci. 7 (1998) 21–38.[34] F.E. Regnier, High-performance liquid chromatography of biopolymers, Science 222

(1983) 245–252.[35] X. Geng, F.E. Regnier, Retentionmodel for proteins in reversed-phase liquid chroma-

tography, J. Chromatogr. 296 (1984) 15–30.[36] X. Geng, F.E. Regnier, Stoichiometric displacement of solvent by non-polar solutes in

reversed-phase liquid chromatography, J. Chromatogr. 332 (1985) 147–168.[37] S.L. Wu, A. Figueroa, B.L. Karger, Protein conformational effects in hydrophobic

interaction chromatography. Retention characterization and the role of mobilephase additives and stationary phase hydrophobicity, J. Chromatogr. 371 (1986)3–27.

[38] B.L. Karger, R. Blanco, The effect of on-column structural changes of proteins on theirHPLC behavior, Talanta 36 (1989) 243–248.

[39] N.E. Zhou, C.T. Mant, R.S. Hodges, Effect of preferred binding domains on peptide re-tention behavior in reversed-phase chromatography: amphipathic alpha-helices,Pept. Res. 3 (1990) 8–20.

[40] S.E. Blondelle, J.M. Ostresh, R.A. Houghten, E. Perez-Paya, Induced conformationalstates of amphipathic peptides in aqueous/lipid environments, Biophys. J. 68(1995) 351–359.

[41] A.W. Purcell, M.I. Aguilar, R.E. Wettenhall, M.T. Hearn, Induction of amphipathichelical peptide structures in RP-HPLC, Pept. Res. 8 (1995) 160–170.

[42] C.T. Mant, J.M. Kovacs, H.M. Kim, D.D. Pollock, R.S. Hodges, Intrinsic amino acid side-chain hydrophilicity/hydrophobicity coefficients determined by reversed-phasehigh-performance liquid chromatography of model peptides: comparison withother hydrophilicity/hydrophobicity scales, Biopolymers 92 (2009) 573–595.