RESEARCH PAPER

An analytical workflow for the molecular dissectionof irreversibly modified fluorescent proteins

Vivien Berthelot & Vincent Steinmetz & Luis A. Alvarez &

Chantal Houée-Levin & Fabienne Merola &

Filippo Rusconi & Marie Erard

Received: 7 June 2013 /Revised: 8 July 2013 /Accepted: 11 July 2013 /Published online: 12 September 2013# Springer-Verlag Berlin Heidelberg 2013

Abstract Owing to their ability to be genetically expressed inlive cells, fluorescent proteins have become indispensablemarkers in cellular and biochemical studies. These proteinscan undergo a number of covalent chemical modifications thatmay affect their photophysical properties. Among other mech-anisms, such covalent modifications may be induced by reac-tive oxygen species (ROS), as generated along a variety ofbiological pathways or through the action of ionizing radia-tions. In a previous report [1], we showed that the exposure ofcyan fluorescent protein (ECFP) to amounts of •OH thatmimic the conditions of intracellular oxidative bursts (associ-ated with intense ROS production) leads to observablechanges in its photophysical properties in the absence of anydirect oxidation of the ECFP chromophore. In the presentwork, we analyzed the associated structural modifications ofthe protein in depth. Following the quantified production of•OH, we devised a complete analytical workflow based onchromatography and mass spectrometry that allowed us tofully characterize the oxidation events. While methionine,tyrosine, and phenylalanine were the only amino acids that

were found to be oxidized, semi-quantitative assessment oftheir oxidation levels showed that the protein is preferentiallyoxidized at eight residue positions. To account for the pre-ferred oxidation of a few, poorly accessible methionine resi-dues, we propose a multi-step reaction pathway supported bydata from pulsed radiolysis experiments. The described ex-perimental workflow is widely generalizable to other fluores-cent proteins, and opens the door to the identification ofcrucial covalent modifications that affect their photophysics.

Keywords Protein oxidation .Mass spectrometry . •OHradicals . Cyan fluorescent protein (ECFP) . γ- and pulsedradiolysis

Introduction

Fluorescent proteins (called “FPs” for short) have revolution-ized mechanistic investigations of cellular processes by livecell imaging, flow cytometry, and high-throughput bioassays[2]. FPs comprise 11 β-sheets that form a barrel enclosing acoaxial α-helix (Fig. 1a); this α-helix bears the chromophoreresulting from the cyclization of three consecutive aminoacids [65–67] (Fig. 1b). As such, these proteins have beenused to build fully gene-encoded sensors that allow a large setof biological chemical events to be monitored in live cells [3].

Engaging in chemical sensing through the use of FPsrequires a thorough knowledge of their photophysical re-sponses to the cellular microenvironment. While some atten-tion has been paid to the side effects of either pH [4–7],chloride ions [8], or refractive index [9] on the photophysicalproperties of FPs, the possible consequences of the localproduction of reactive oxygen species (ROS: O2

• −, •OH,H2O2, HOCl) have seldom been addressed [1, 10–12]. Nev-ertheless, ROS play crucial roles in physiological and patho-logical processes: they can be produced in large quantities in

F. Rusconi and M. Erard are both corresponding authors of this work.

Electronic supplementary material The online version of this article(doi:10.1007/s00216-013-7326-y) contains supplementary material,which is available to authorized users.

V. Berthelot :V. Steinmetz : L. A. Alvarez : C. Houée-Levin :F. Merola : F. Rusconi (*) :M. Erard (*)Laboratoire de Chimie Physique, UMR CNRS 8000, Building 350,91405 Orsay Cedex, Francee-mail: filippo.rusconi@u-psud.fre-mail: marie.erard@u-psud.fr

F. RusconiRégulation et Dynamique des Génomes, U INSERM 565—UMRCNRS 7196, Muséum National d’Histoire Naturelle, 57 rue Cuvier,Case Postale 26, 75231 Paris Cedex 05, France

Anal Bioanal Chem (2013) 405:8789–8798DOI 10.1007/s00216-013-7326-y

cells, and are potentially harmful as they cause oxidativedamage—so-called oxidative stress [13, 14]. For example,the dynamics of pathogen phagocytosis, a phenomenon cor-related with high levels of ROS production by enzymaticsystems such as NADPH oxidase or myeloperoxidase [15],was recently investigated with the aid of FPs [16–19], whilethere has been a recent push to develop members of the FPfamily as specific oxidation sensors [10, 11, 18, 20–22]. Insome of the latter cases [10, 11, 18, 22], as well as in otherreports [1, 12], the consequences of ROS exposure for thephotophysical properties of FPs have been reported, and theseconsequences were shown to be dependent on either thenature and quantity of the oxidant or the FP variant employed.On the other hand, the characterization of ROS-induced chem-ical modifications of FPs may prove fundamental to advanc-ing our understanding of their photobleaching mechanisms.Indeed, these photoreactions are thought to result from theuncontrolled production of oxidants (like ROS) from singletor triplet excited states of the fluorophore [23]. Photobleachinggenerates variations in the fluorescence signal, known asphotofatigue, thus limiting the duration and the reliability ofthe measurements. This photophysical limitation becomes crit-ical when high illumination densities are required, as in super-resolution optical microscopy or single-molecule applications[24, 25]. So far, very few studies have been devoted to theidentification of chemical modifications involved inphotobleaching [23, 26], and both the pathways and the targetsare generally still unknown. In a previous study, we analyzed,in detail, the photophysical perturbations induced by controlledamounts of hydroxyl (•OH) and superoxide anion (O2

• −) freeradicals on the cyan fluorescent protein (ECFP) [1], which isthe most widely used donor in FRET-based imaging experi-ments in combination with a yellow partner (e.g., citrine) [3,27]. We found significant perturbations of the ECFP fluores-cence (decreases in intensity and lifetime, without changes inits excitation and emission spectra) that did not result fromdirect chemical modification of the chromophore but ratherfrom oxidations at other sites that are still to be identified.

Understanding the causative relationships between the ex-posure of FPs to ROS and the resulting modifications of theirphotophysical properties requires thorough structural charac-terization of the oxidized proteins. Due to the lack of appro-priate analytical tools, these questions have not yet beenaddressed in detail. Crystallography may appear to be anappropriate approach because fluorescent proteins are highlycompact and can be crystallized easily. However, the expectedmultiplicity of the chemical modifications limits the use ofcrystallography in this context. The bottom-up strategy that isoften used in proteomics should holdmore promise. However,the compactness of the FP barrel structure makes these pro-teins refractory to conventional enzymatic breakdown, andthus previous mass spectrometric analyses (briefly reviewedby Alvarez et al. [28]) have had to resort to chemically harsh

procedures, which risked causing unwanted chemical modifi-cations of the proteins. In a previous work, we devised a newmethod of digesting FPs in very mild conditions that openedthe door to the molecular dissection of chemically modifiedFP variants [28].

In the work described in the present paper, we built on thatprevious work and further enhanced our analytical workflowto achieve the full characterization of •OH-oxidized ECFP.This led to the identification of ECFP-oxidized residues, alongwith the first semi-quantitation of their oxidation events. Wewere able to spot oxidation events, which were mostly restrict-ed to a few specific residues that may be responsible for theobserved photophysical changes. In addition, pulsed radioly-sis studies provided further insights into the primary targets of•OH that lead to these selective oxidations.

Experimental section

Cyan fluorescent protein purification and endoAspN-baseddigestion

ECFP refers to AvGFP F64L/S65T/Y66W/N146I/M153T/V163A/H231L (Clontech, Mountain View, CA, USA). Theproduction and the purification of His-tagged recombinantECFP were performed as described previously [1]. The Histag was cleaved away, leading to a cyan fluorescent protein inwhich the starting M residue is replaced by the GA dipeptide.The residue numbering scheme relies on the presence of thechromophore TWG triplet residues at positions 65-66-67 (seeFig. 1b). The ECFP solution was dialyzed against a 30-mMphosphate buffer at pH 7.5 for the irradiation experiments.The protein concentration was assessed by UV absorption ofthe chromophore [ε(λ434nm)=30,000 cm−1 M−1].

Radical production and radiolysis

A detailed description of the radiolysis procedure is provided insection OR1 of the “Electronic supplementary material”(ESM). Briefly, the quantitative production of •OH wasperformed using either stationary γ-radiolysis or high-energyelectron pulses. They were selected using the well-knownmethod employing scavengers in solutions deoxygenated andsaturated with N2O [29]. Additional pulsed radiolysis experi-ments were performed using •N3 radicals as oxidants. For theirradiations, the ECFP concentration was usually 5 μM (unlessotherwise specified), at neutral pH. The doses applied were inthe range 20–400 Gy (stationary radiolysis) or in the range 4–10 Gy (pulsed radiolysis). To make it easier to understand theresults, the dose was replaced with the ratio R =[•OH]/[ECFP],which is proportional to the irradiation dose.

The pulsed-radiolysis-elicited reaction was followed spec-trophotometrically between λ 350nm and λ 750nm, and the

8790 V. Berthelot et al.

difference absorption spectra of the reaction intermediateswere reconstructed from the maxima of the absorbance vari-ation in the investigated wavelength range.

Structural modeling of residue accessibility

The structure of ECFP at physiological pH is available fromPDB (ID: 2WSN) [30], but does not include the C-terminaltail of the protein (sequence 230 TLGMDELYK 238). In orderto correlate the mass spectrometric results with the spatiallocations of all of the ECFP residues, we generated an auto-mated model of ECFP using the SWISS-MODEL workspacedescribed in section OR2 of the ESM.

We calculated the solvent-accessible surface percentage ofeach residue using the GETAREA service provided by theSealy Center for Structural Biology at the University of TexasMedical Branch (Galveston, TX, USA) [31].

Chromatography

Desalting of the proteins and peptide mixtures was performedaccording to the procedure described in [32], with the R2Poros polymeric resin (Applied Biosystems, Carlsbad, CA,USA) employed as the reversed phase.

Peptides from endoAspN-digested ECFP [28] were sepa-rated by reversed-phase high-performance liquid chromatog-raphy (RP-HPLC) on an Åkta purifier setup from GEHealthcare (Amersham, UK), using a Stability S-C23 end-capped resin (Cil Cluzeau, Sainte-Foy-la-Grande, France).The gradient was developed from 100 % buffer A [water,0.035 % trifluoroacetic acid (TFA)] to 100 % buffer B[95 % acetonitrile in 0.035 % TFA] in 60 min. Each fractionwas collected manually as the peak shape was monitored on

the chromatogram view in the Unicorn software package (GEHealthcare).

Mass spectrometry

Fluorescent proteins (irradiated or not) or peptides thereofwere analyzed by electrospray ionization (using either normal,microspray, or nanospray) on three different mass spectrome-ters: (1) a QStar Pulsari hybrid quadrupole–time-of-flight(Applied Biosystems); (2) a 7-T Apex Qe FT-ICR with aquadrupole–hexapole interface allowing for m /z -based selec-tion, ion accumulation, and cooling before injection into theICR cell (Bruker Daltonics, Bremen, Germany); (3) an LTQOrbitrap (Thermo Fisher Scientific,Waltham,MA, USA)witha microspray source in-line with a Dionex (Sunnyvale, CA,USA) Ultimate 3000 HPLC setup. In each case, the typicalionization protocol was used, with standard settings that wereappropriate for either an∼30 kDa protein or peptides. All theexperiments were performed in the positive ionization mode.The typical amount of protein used for one analysis was 25μLof a 10-μM solution. The peptidic solutions were typically inthe 2–5 μM range. When peptide analyses were performed byMALDI-TOF mass spectrometry, the instrument used was a4800 MALDI TOF/TOF analyzer from AB Sciex (Framing-ham, MA, USA), and the sample was prepared according tothe conventional dried droplet procedure using α-cyano-4-hydroxycinnamic acid (ACHCA) as the matrix. All gas-phase fragmentation spectra were acquired using collision-induced dissociation (CID). The CIDmass data were analyzedto locate the oxidation events of m /z -selected ions.

Mass spectrometric data analysis was performed by firstexporting the data to simple ASCII-formatted files whichcould then be used by free and open-source software in allof the analytical steps. Mass spectrum visualization and

Fig. 1 a Barrel structure of ECFP. b ECFP sequence. Residues that were found to be oxidized are denoted using filled or semi-filled circles (see text fordetails). The chromophore is typeset in bold blue characters

Molecular dissection of modified fluorescent proteins 8791

analysis was performed using the mMass program (http://mmass.org) [33, 34]. Mass spectrometric data simulationsand detailed analyses were performed using either themassXpert program (http://massxpert.org) [35, 36] or GNUpolyXmass [37]. Data handling was performed by setting up arelational database using the free and open-source SQLitedatabase (http://sqlite.org). All in-house programming wasperformed using either C++ or Python as the programminglanguage on a Debian GNU/Linux computing platform (http://debian.org).

Results and discussion

Oxidation of the ECFP by •OH

After performing γ-irradiation of the purified ECFP, with thedelivered dose yielding a ratio R =[•OH]/[ECFP] of 10, theprotein was desalted/concentrated and analyzed by ESI-MS.Figure 2 shows the spectra obtained for nonirradiated andirradiated ECFP.

The unoxidized protein (panel a) produced a charge enve-lope with peaks that had the expected m /z ratio (e.g., mea-sured: m /z 841.21, z =32; calculated: m /z 841.01). Afteroxidation with •OH (panel b), that same peak almostdisappeared and several new peaks appeared (labeled withasterisks), which were separated from each other by m /z 0.5,corresponding to the binding of oxygen atoms to the protein(ΔM =16 u). At least five resolved oxidation levels wereobserved. Under our experimental conditions, the γ-radiolysis of water yields a non-negligible amount of •Hradicals ([•H]/[ECFP]≈1), but we did not detect any of theirreaction products, such as losses of H2S (ΔM =−34 u) orCH3SH (ΔM =−48 u) [38–41]. Further, at the whole-proteinlevel, no other •OH-induced modification (such as decarbox-ylation or intermolecular Cys- or Tyr-based dimerization; for areview of ROS-induced chemical modifications, see [42]) wasobserved.

A dose–response experiment was performed wherebyECFP was submitted to •OH oxidation with R ratios in therange 0–17. There was a clear correlation between increasingR ratio and the appearance of ECFP-oxidized variants, and aconcomitant decrease in the mass peak corresponding to theunoxidized ECFP, as shown in Fig. 2c. With an R ratio of 17,the protein was so heavily modified that the mass spectrumobtained was uninformative: the molecular diversity associat-ed with the large number of modified ECFP polypeptides ledto the classical spectral suppression phenomenon. On theother hand, for oxidation ratios of ∼17, we observed only a15 % decrease in both the fluorescence quantum yieldand lifetime of ECFP [1]. Such moderate concomitantphotophysical perturbations indicate that many of these ECFP

oxidations actually have little or no impact on the protein’sphotophysical properties.

Fig. 2a–c Mass spectrometry-based monitoring of the oxidation of ECFPby •OH. ECFP was either unoxidized (a) or oxidized by •OH with R=10(b). The inset shows a close-up of the mass peak (z=32) enclosed by thedashed box . The unoxidized protein was almost homogeneous, while the•OH-oxidized protein showed several variants differing by the mass of oneormore oxygen atoms (ΔM=16 u). c Progression of the oxidation of ECFPas observed upon increasing the R=[•OH]/[ECFP] ratio (numbers next tothe traces). The most useful R values for our experiments were found to liein the range 4–10. Arrows point to the mass peak corresponding to theunmodified polypeptide

8792 V. Berthelot et al.

Characterization of the •OH-oxidized variants of ECFP

Direct analysis of the oxidized ECFP endoAspN-producedpeptidic mixture

Oxidized ECFP samples were subjected to endoAspN diges-tion according to the protocol described in [28]. The peptidicmixture was analyzed by ESI mass spectrometry, using flowinjection, direct infusion, or a nanospray. These mass spectro-metric analyses failed to yield the correct sequence coverage,which contrasted with the excellent coverage for unoxidized

ECFP that we described in an earlier report [28]. Most evidentwas the complete lack of a signal corresponding to thechromopeptide (coordinates [36–75]) comprising the three65 TWG 67 residues that undergo a series of post-translational modifications to form the chromophore (repre-sented using bold blue characters in Fig. 1b). This peptide isone of the largest peptides expected from ECFP digestion byendoAspN, even with partial cleavages occurring in otherregions of the protein sequence. We reasoned that this insuf-ficient sequence coverage was due to a huge increase in themolecular complexity of the peptidic mixture in the case ofoxidized ECFP. Indeed, the whole set of oxidation combina-tions for each oxidizable peptide may have made the sampletoo complex to be successfully analyzed without sufferingfrom spectral suppression. Furthermore, because the oxidationof peptides increases their hydrophilicity, the desolvation/ionization of the oxidized peptides could be disfavored com-pared to the desolvation/ionization of their unmodified coun-terparts [43], thus leading to a loss of mass spectrometricsignal that would mainly affect the subpopulation of peptidesthat are of greatest interest in this study [44].

Chromatographic separation of ECFP peptides

In a first attempt to reduce mass spectral suppression, amicrochromatography experiment was performed as de-scribed in [32]. Step gradient elution was carried out, withtwo (20 % and 60 %) acetonitrile fractions collected. Analysisof the fractions showed that a substantial proportion of themass spectrometric signal could be recovered. Indeed, the

Fig. 3 Reversed-phase high-performance liquid chromatography sepa-ration of •OH-oxidized ECFP peptides following endoAspN digestion.Absorbance was measured at the wavelengths λ214nm and λ414nm in orderto detect the peptidic bond and the chromophore, respectively. Theamount of protein injected was 200 pmol. Each fraction was collectedin a separate tube for further analysis by mass spectrometry

Table 1 ECFP-oxidized peptidedata for R =[•OH]/[ECFP] in therange 4–10

SAA surface-accessible area. Theoxidized residues were classifiedinto four categories: those thatwere found to be unoxidized (∅),barely oxidized (<2–5 %; ○),weakly oxidized (<10–20 %; ◑),and strongly oxidized (>20–40%;●). A question mark indicates thatno oxidized residue was found

ID Sequence Coordinates Oxidized residue/abundance SAA %

1 GAVSKGEELFTGVVPILVEL [−18] ?/○ –

2 DVNGHKFSVSGEGEG [21–35] F27/◑ 0.7

3 DHMKQH [76–81] M78/● 0.9

4 DFFKSAMPEGYVQERTIFFK [82–101] F83/∅ 0.5

F84/∅ 0.2

M88/● 0

Y92/∅ 0

F99/∅ 58.3

F100/∅ 0.2

5 DGNYKTRAEVKF [103–114] ?/○ –

6 EYNYISHNVYITA [142–154] ?/○ –

Y151/◑ 59.6

7 DGPVLLPDNHYLSTQSALSK [190–209] Y200/◑ 33.0

8 DHMVLLEFVTAAGITLGMDELYK [216–238] M218/◑ 0

F223/∅ 40.2

M233/● 58.4

Y237/● 100

Molecular dissection of modified fluorescent proteins 8793

chromopeptide was detected in the 60 % acetonitrile fraction;however, sequence coverage was still only partial.

The peptidic mixture obtained upon endoAspN digestionof ECFP was thus subjected to an HPLC separation using areversed-phase resin. Figure 3 shows the chromatogramobtained, with the absorbance measured at λ 214nm andλ414nm. The latter wavelength corresponds to the absorptionband of the chromopeptide [4], thus allowing us tomonitor thefraction in which it eluted (at 56min). All of the fractions werecollected separately and later analyzed by ESI or MALDImass spectrometry to identify the oxidized peptides. Oneuseful observation was that oxidized peptides almost system-atically eluted in the fraction preceding that containing theirunoxidized peptide counterparts. This observation was bothexpected (because the oxidized peptides are more hydrophilic)and helpful when analyzing the mass spectrometric data,because it made it possible to predict the fraction that shouldbe searched for oxidized variants of any given peptide.

Full mapping of the oxidation sites in ECFP

The primary structure of ECFP was completely covered byour mass spectrometric analyses. In agreement with our pre-vious findings (Fig. 2), we focused our data analyses onoxygen additions (ΔM =16 u). Entirely manual scrutiny ofthe mass spectrometric data obtained for the oxidized ECFP-derived peptides allowed us to map the oxidation events to afew positions in the protein sequence. The residues that werefound to be oxidized were all either aromatic or sulfur-containing residues [45, 46]. Two kinds of residues wereobserved: residues oxidized with R =[•OH]/[ECFP] in therange 4–10, and residues that only occurred as oxidized spe-cies upon intense irradiation (R ≈20). In Table 1 and Fig. 1b,we list only the peptides that were oxidized with R in therange 4–10. For each listed peptide, the residue that was found

to undergo the oxidation is specified, along with its abundanceand its solvent accessibility. The different residues belongingto the oxidized peptides were classified into four categories:those that were found to be unoxidized (∅), barely oxidized(<2–5 %; ○), weakly oxidized (<10–20 %; ◑), and stronglyoxidized (>20–40%; ●). The amount of each oxidized peptiderelative to its unoxidized counterpart in the same sample wasassessed by comparing the peak intensities for both forms; thiscomparison was expressed as the ratio (given as a percentage)of oxidized to unoxidized mass peak intensity.

Three of the four methionine residues were strongly oxi-dized into methionine sulfoxide (M78, M88, and M233),while the percentage of M218 sulfoxide was lower. We didnot find any methionine sulfone [42]. Among the ten tyrosineresidues, only three were found to be significantly oxidized toDOPA (Y151, Y200, and Y237). Interestingly, none of these

Fig. 4 Mass spectrometric analysis of the •OH-oxidized ECFPchromopeptide (R ≈20). Isotopic clusters were found to correspond tothe chromopeptide, both in the unoxidized form (at m /z 1120.83, z =4)and in the singly oxidized form (atm /z 1124.85). The chromopeptide wasfound to have a disulfide bond

Fig. 5 a Absorbance (at λ315nm and λ390nm) traces recorded after thepulse. Experimental conditions were, for λ315nm, [ECFP]=30 μM,[•OH]=7 μM; for λ390nm, curve 1: [ECFP]=30 μM, [•OH]=7 μM; forλ390nm, curve 2: [ECFP]=12 μM, [•OH]=4.5 μM. b Difference absorp-tion spectra recorded 40 μs and 500 μs after the pulse for •OH radicalsand •N3 radicals, respectively. Experimental conditions were [ECFP]=12 μM, [•OH]=4.5 μM. The time response of the experimental setup wasbelow 0.1 μs

8794 V. Berthelot et al.

tyrosine residues underwent oxidation to bityrosine (anotherfrequent product of tyrosine oxidization [42, 47])—neitherintermolecular [1] nor intramolecular. In the latter case, theonly intramolecular bityrosine that could have formed wouldhave involved Y151 and Y200 (d =4.8 Å). In that case, thecreation of a covalent bond between peptides 6 and 7 wouldhave changed the masses of the resulting endoAspN-producedpeptides. This was checked, and the mass corresponding to thelinked peptide was not observed in the irradiated samples. Inaddition, F27 was found to be oxidized, as seen in otherpeptides ([48] and Ignasiak et al., personal communication).Oxidation events for barely oxidized peptides could not bemapped because of the low precursor ion signal, which im-peded their gas-phase fragmentation sequencing.

Upon high [•OH]/[ECFP] oxidation (R ≈20), the most no-table difference was the appearance of an oxidized variant ofthe chromopeptide, as demonstrated in Fig. 4, which showsthe isotopic cluster obtained by ESI–FT-ICR mass spectro-metric analysis of the chromatographic fraction absorbing atλ 414nm (measured: m /z 1120.83, z =4; calculated: m /z1120.84). This mass peak corresponded to the chromopeptidethat contained an intact TWG chromophore and in which aCys48–Cys70 disulfide bond had formed. This disulfide bondcould only have formed after the endoAspN hydrolysis of theprotein, because the cysteine residues in the native ECFPbarrel structure are too far apart to form a bond [28]. Onepeak at m /z 1124.85 was found to be at Δm /z =+4 from theprevious one, which corresponded to the [M+4H+O]4+

oxidation variant of the chromopeptide. However, our previ-ous photophysical analysis of the •OH-oxidized ECFPallowed us to exclude the oxidations of W57 and W66 up toR =45 (W66 belongs to the chromophore) [1].

To summarize, the only oxidation events that could beobserved were oxygen additions outside the ECFP chromo-phore. The detailed analysis of the addition sites actuallyyielded a few targets (F, M, and Y, labeled ◑ and ● in Fig. 1and Table 1). It is worth noting that a few of the residues thatwe found in an oxidized form were buried inside the ECFPbarrel structure (their surface-accessible areas were below1 %; Table 1). Because of the high chemical reactivity of•OH, the presence of buried oxidized residues hinted at indi-rect oxidation mechanisms, which we then investigated byperforming the pulsed radiolysis experiments described next.

Pulsed radiolysis

The first steps in the reaction of •OH with ECFP were inves-tigated by pulsed radiolysis, where the timespan for •OHcreation is short compared to the reaction time. This methodallowed the identification of reaction intermediates by theirtransient absorption spectra, and permitted us to probe theirformation and decay kinetics.

The absorbance trace (which was monitored to determine thekinetics) recorded after the pulsed production of •OH radicalsvaried according to the wavelength observed (Fig. 5a). At

Fig. 6 Reactions proposed to explain the obtained kinetics traces

Molecular dissection of modified fluorescent proteins 8795

λ315nm, the absorbance reached its maximum a few micro-seconds after the pulse and then decayed. The build-upphase corresponded to a rate constant of (2.5±0.5) ×1010 mol−1 L s−1, which is consistent with that of the reactionof •OHwith proteins of the same size [49]. In the range λ390nm

−410nm, the absorbance reached its maximum ca. 50 μs afterthe pulse and then decayed slowly in∼100 ms (not shown).These wavelength-dependent kinetics show that several—chemically different—intermediates are produced in the firstfew tens of microseconds after the pulsed •OH production.Moreover, the timescale of the absorption build-up at λ390nm

did not depend on the [ECFP] nor on the [•OH] (Fig. 5a),indicating an intramolecular process.

The time-resolved absorption spectrum of the ECFP solu-tion recorded 40 μs after the •OH pulse is shown in Fig. 5b(•OH trace). The signal at λ320nm can be attributed to theaddition of •OH to Tyr and Phe [50, 51]. The bands in theλ390nm−410nm range can be attributed to ECFP–(TyrO•) and/orto ECFP–(MetS ∴ X)•+ (X=O, N, or S) [45, 46, 52, 53](Fig. 5b). To confirm the nature of these radicals, we comparedthis spectrum with the one obtained after oxidation using azideradicals (Fig. 5b, •N3 trace), which are known to participate inelectron transfer with tyrosine; they therefore yield only theECFP–(TyrO•) radical, and exhibit low reactivity with methio-nine [54, 55]. The new spectrum was different from the previ-ous one, with shifted maxima, confirming that TyrO-centeredradicals are not the only short-lived products of the ECFPoxidation, and that methionine-centered radicals are present.

In proteins, the tyrosine residue is usually thought to be theendpoint of oxidation. For example, oxidations of tryptophanor methionine residues would, in our case, end up with theformation of ECFP–(TyrO•) by intramolecular electron trans-fer [56]. However, in our experiments, we found that severalmethionine residues are ultimately oxidized, despite the factthat some of these residues are deeply buried in the proteinstructure. In addition, the reaction leading to the increasedabsorbance in the λ390nm region in pulsed radiolysis experi-ments was first order, indicating an intramolecular reaction.Figure 6 shows our proposed interpretation of these observa-tions. •OH radicals can react with methionine or tyrosineresidues, as observed in the transient absorption spectra shownin Fig. 5 (steps a and b, respectively). The •OH adduct on thetyrosine residue may lead to DOPA (step c) or undergoproton-catalyzed dehydration to the TyrO• radical (step d),which in turn is repaired by a methionine residue (step e),thus leading to methionine sulfoxide (step f). In this protein,the distances between the methionine and tyrosine residuescan be short (d [M78–Y200]=8.2 Å, d [M218–Y143]=4.6 Åand d [M233–Y237]=5.8 Å) or long (d [M88–Y39]=20 Å;see Fig. S1 of the ESM). In the former case, the electrontransfer should be fast, while in the latter case it can proceedby hopping through aromatic residues [57, 58], as illustrated inFig. S1 of the ESM. In all cases, the electron transfer requires

that the reduction potential of the methionine residue is lowerthan that of tyrosine, which may indeed happen, because themethionine reduction potential is highly dependent on its im-mediate environment [59]. Furthermore, other studies on pep-tides or proteins have shown that many oxidations do notalways end up at tyrosine residues [60–63].

Conclusions

In this work, the described analytical workflow showed thatfluorescent protein oxidation was observed on methionine,tyrosine, and phenylalanine residues. One interesting findingis that three-dimensional modeling of the protein places someof the oxidized residues in the inner face of the barrel, hintingat indirect oxidation pathways involving one or more hopsacross the protein structure. This observation is of particularinterest in the context of studies involving •OH-based oxida-tion of proteins to determine the three-dimensional structureof proteins or the topological organization of protein assem-blies [64, 65]. Our work could indeed point to one difficulty inthis field, which may arise if the ultimate oxidation state ofany given residue is not solely the result of its accessibility tothe solvent.

Another significant result, along with previous results fromour laboratory [1], is that the chemical modifications identi-fied in this report perturb the photophysical properties of thefluorescent protein, even when the chromophore is not mod-ified. This observation is consistent with our previously pro-posed hypothesis that the oxidation of the protein increases theflexibility of the barrel, thus enhancing nonradiative de-excitation paths through excited-state chromophore torsion[1]. In addition, the present work shows that these dynamicalperturbations may be triggered by only a few critical oxida-tions in the ECFP structure, all of which are located at least8 Å away from the chromophore, thus excluding the directquenching of the chromophore.

Acknowledgments V.B. received a doctoral fellowship from the Uni-versity of Paris-Sud, Orsay, France. L.A.A. was supported by a fellow-ship from the French Ministry of Research (MESR). The authors thankDr. Philippe Maître (plateforme de spectrométrie de masse du LCP,University of Orsay) for interesting discussions, Drs. Lionel Dubost andArul Marie of the mass spectrometry facility of the Muséum in Paris,France, and Dr. Jean-Pierre Le Caer of the mass spectrometry facility ofthe ICSN in Gif-sur-Yvette, France. We thank Dr. Vincent Favaudon forthe use of the pulsed radiolysis setup (Institut Curie, Orsay, France). Weare indebted to the COST action CM1001 (non-enzymatic proteinoxidation) for fruitful discussions.

References

1. Alvarez L, Levin CH, Merola F, Bizouarn T, Pasquier H, Baciou L,Rusconi F, Erard M (2010) Are the fluorescent properties of the cyan

8796 V. Berthelot et al.

fluorescent protein sensitive to conditions of oxidative stress?Photochem Photobiol 86(1):55–61

2. Tsien RY (1998) The green fluorescent protein. Annu Rev Biochem67:509–544

3. Newman RH, Fosbrink MD, Zhang J (2011) Genetically encodablefluorescent biosensors for tracking signaling dynamics in living cells.Chem Rev 111(5):3614–3666

4. Fredj A, Pasquier H, Demachy I, Jonasson G, Levy B, Derrien V,Bousmah Y, Manoussaris G, Wien F, Ridard J, Erard M, Merola F(2012) The single T65S mutation generates brighter cyan fluorescentproteins with increased photostability and pH insensitivity. PLoSOne7(11):e49149

5. Villoing A, Ridhoir M, Cinquin B, Erard M, Alvarez L, Vallverdu G,Pernot P, Grailhe R, Merola F, Pasquier H (2008) Complex fluores-cence of the cyan fluorescent protein: comparisons with the H148Dvariant and consequences for quantitative cell imaging. Biochemistry47(47):12483–12492

6. ErardM, Fredj A, Pasquier H, Beltolngar DB, BousmahY, Derrien V,Vincent P, Merola F (2013) Minimum set of mutations needed tooptimize cyan fluorescent proteins for live cell imaging. Mol Biosyst9(2):258–267

7. Poëa-Guyon S, Pasquier H, Merola F, Morel N, Erard M (2013) Theenhanced cyan fluorescent protein: a sensitive pH sensor for fluores-cence lifetime imaging. Anal Bioanal Chem 405(12):3983–3987

8. GriesbeckO, Baird GS, Campbell RE, Zacharias DA, Tsien RY (2001)Reducing the environmental sensitivity of yellow fluorescent protein.Mechanism and applications. J Biol Chem 276(31):29188–29194

9. van Manen HJ, Verkuijlen P, Wittendorp P, Subramaniam V, van denBerg KT, Roos D, Otto C (2008) Refractive index sensing of greenfluorescent proteins in living cells using fluorescence lifetime imag-ing microscopy. Biophys J 94(8):L67–L69

10. Palazzolo AM, Suquet C, Konkel ME, Hurst JK (2005) Greenfluorescent protein-expressing Escherichia coli as a selective probefor HOCl generation within neutrophils. Biochemistry 44(18):6910–6919

11. Tsourkas A, Newton G, Perez JM, Basilion JP, Weissleder R (2005)Detection of peroxidase/H2O2-mediated oxidation with enhancedyellow fluorescent protein. Anal Chem 77(9):2862–2867

12. Bou-Abdallah F, Chasteen ND, Lesser MP (2006) Quenching ofsuperoxide radicals by green fluorescent protein. Biochim BiophysActa 1760(11):1690–1695

13. Brieger K, Schiavone S, FJM Jr, Krause KH (2012) Reactive oxygenspecies: from health to disease. Swiss Med Wkly 142:w13659

14. Winterbourn CC (2008) Reconciling the chemistry and biology ofreactive oxygen species. Nat Chem Biol 4(5):278–286

15. Winterbourn CC, Hampton MB, Livesey JH, Kettle AJ (2006)Modeling the reactions of superoxide and myeloperoxidase in theneutrophil phagosome: implications for microbial killing. J BiolChem 281(52):39860–39869

16. Hoppe AD, Swanson JA (2004) Cdc42, Rac1, and Rac2 displaydistinct patterns of activation during phagocytosis. Mol Biol Cell15(8):3509–3519

17. Ueyama T, Tatsuno T, Kawasaki T, Tsujibe S, Shirai Y, Sumimoto H,Leto TL, Saito N (2007) A regulated adaptor function of p40phox:Distinct p67phox membrane targeting by p40phox and by p47phox.Mol Biol Cell 18(2):441–454

18. Schwartz J, Leidal KG, Femling JK, Weiss JP, Nauseef WM (2009)Neutrophil bleaching of GFP-expressing staphylococci: probing theintraphagosomal fate of individual bacteria. J Immunol 183(4):2632–2641

19. Tlili A, Erard M, Faure MC, Baudin X, Piolot T, Dupre-Crochet S,Nusse O (2012) Stable accumulation of p67phox at the phagosomalmembrane and ROS production within the phagosome. J LeukocBiol 91(1):83–95

20. Belousov VV, Fradkov AF, Lukyanov KA, Staroverov DB,Shakhbazov KS, Terskikh AV, Lukyanov S (2006) Genetically

encoded fluorescent indicator for intracellular hydrogen peroxide.Nat Methods 3(4):281–286

21. Meyer AJ, Dick TP (2010) Fluorescent protein-based redox probes.Antioxid Redox Signal 13(5):621–650

22. Wang W, Fang H, Groom L, Cheng A, Zhang W, Liu J, Wang X, LiK, Han P, Zheng M, Yin J, Wang W, Mattson MP, Kao JPY, LakattaEG, Sheu SS, Ouyang K, Chen J, Dirksen RT, Cheng H (2008)Superoxide flashes in single mitochondria. Cell 134(2):279–290

23. Bourgeois D, Regis-Faro A, Adam V (2012) Photoactivated structuraldynamics of fluorescent proteins. Biochem Soc Trans 40(3):531–538

24. Lippincott-Schwartz J, Patterson GH (2009) Photoactivatable fluo-rescent proteins for diffraction-limited and super-resolution imaging.Trends Cell Biol 19(11):555–565

25. Ha T, Tinnefeld P (2012) Photophysics of fluorescent probes forsingle-molecule biophysics and super-resolution imaging. AnnuRev Phys Chem 63:595–617

26. van Thor JJ, Gensch T, Hellingwerf KJ, Johnson LN (2002)Phototransformation of green fluorescent protein with UVand visiblelight leads to decarboxylation of glutamate 222. Nat Struct Biol 9(1):37–41

27. Piston DW, Kremers GJ (2007) Fluorescent protein FRET: the good,the bad and the ugly. Trends Biochem Sci 32(9):407–414

28. Alvarez LA, Merola F, Erard M, Rusconi F (2009) Massspectrometry-based structural dissection of fluorescent proteins. Bio-chemistry 48(18):3810–3812

29. Spinks JWT, Woods RJ (1990) An introduction to radiation chemis-try. Wiley, New York

30. Lelimousin M, Noirclerc-Savoye M, Lazareno-Saez C, Paetzold B,Vot SL, Chazal R, Macheboeuf P, Field MJ, Bourgeois D, Royant A(2009) Intrinsic dynamics in ECFP and Cerulean control fluores-cence quantum yield. Biochemistry 48(42):10038–10046

31. Fraczkiewicz R, Braun W (1998) Exact and efficient analyticalcalculation of the accessible surface areas and their gradients formacromolecules. J Comp Chem 19:319–333

32. Rusconi F, Schmitter JM, Rossier J, le Maire M (1998) Chromato-graphic separation and sample preparation in one step for MALDImass spectrometric analysis of subpicomole amounts of heteroge-neous protein samples. Anal Chem 70(14):3046–3052

33. Strohalm M, Hassman M, Kosata B, Kodicek M (2008) MMass dataminer: an open source alternative for mass spectrometric data analy-sis. Rapid Commun Mass Spectrom 22(6):905–908

34. Strohalm M, Kavan D, Novak P, Volny M, Havlicek V (2010)MMass 3: a cross-platform software environment for precise analysisof mass spectrometric data. Anal Chem 82(11):4648–4651

35. Rusconi F, Belghazi M (2002) Desktop prediction/analysis of massspectrometric data in proteomic projects by using massXpert. Bioin-formatics 18(4):644–645

36. Rusconi F (2009) MassXpert 2: a cross-platform software environ-ment for polymer chemistry modelling and simulation/analysis ofmass spectrometric data. Bioinformatics 25(20):2741–2742

37. Rusconi F (2006) GNU polyxmass: a software framework for massspectrometric simulations of linear (bio-)polymeric analytes. BMCBioinforma 7:226

38. Kadlcik V, Sicard-Roselli C, Houée-Levin C, Kodicek M, Ferreri C,Chatgilialoglu C (2006) Reductive modification of a methionineresidue in the amyloid-beta peptide. Angew Chem Int Ed Engl45(16):2595–2598

39. Mozziconacci O, Bobrowski K, Ferreri C, Chatgilialoglu C (2007)Reactions of hydrogen atoms with met-enkephalin and related pep-tides. Chemistry 13(7):2029–2033

40. Chatgilialoglu C, Ferreri C, Torreggiani A, Salzano AM, Renzone G,Scaloni A (2011) Radiation-induced reductive modifications ofsulfur-containing amino acids within peptides and proteins. J Prote-omics 74(11):2264–2273

41. Torreggiani A, Chatgilialoglu C, Ferreri C, Melchiorre M, Atrian S,Capdevila M (2013) Non-enzymatic modifications in metallothioneins

Molecular dissection of modified fluorescent proteins 8797

connected to lipid membrane damages: structural and biomimetic stud-ies under reductive radical stress. J Proteomics. doi:10.1016/j.jprot.2013.02.005

42. Bachi A, Dalle-Donne I, Scaloni A (2013) Redox proteomics: chem-ical principles, methodological approaches and biological/biomedicalpromises. Chem Rev 113(1):596–698

43. Tang L, Kebarle P (1993) Dependence of ion intensity in electrospraymass spectrometry on the concentration of the analytes in theelectrosprayed solution. Anal Chem 65(24):3654–3668

44. Wilm M (2011) Principles of electrospray ionization. Mol Cell Pro-teomics 10(7):M111–M009407

45. Houée-Levin C, Sicard-Roselli C (2001) Radiation chemistry ofproteins. In: Jonah C, Rao BM (ed) Radiation chemistry: presentstatus and future prospects. Elsevier, Amsterdam

46. Davies MJ, Dean RT (1997) Radical‐Mediated Protein Oxidation,from Chemistry to Medicine. Oxford University Press, Oxford

47. Blouquit Y, Duchambon P, Brun E, Marco S, Rusconi F, Sicard-Roselli C (2007) High sensitivity of human centrin 2 towardradiolytical oxidation: C-terminal tyrosinyl residue as the main target.Free Radic Biol Med 43(2):216–228

48. Mozziconacci O, Mirkowski J, Rusconi F, Kciuk G, Wisniowski PB,Bobrowski K, Houée-Levin C (2012) Methionine residue acts as aprooxidant in the *OH-induced oxidation of enkephalins. J PhysChem B 116(41):12460–12472

49. Buxton GV, Greenstock CL, Helman WP, Ross AB (1988) Critical-review of rate constants for reactions of hydrated electrons,hydrogen-atoms and hydroxyl radicals (OH/O) in aqueous-solution.J Phys Chem Ref Data 17:513–886

50. Lichtin NN, Shafferman A (1974) Reactions of phenylalanine deriva-tives, methionine derivatives, and their mixtures in aqueous solutionwith pulse-radiolytically generated H atoms. Radiat Res 60(3):432–440

51. Solar S, Solar W, Getoff N (1984) Reactivity of hydroxyl withtyrosine in aqueous solution studied by pulse radiolysis. J Phys Chem88(10):2091–2095

52. Holcman J, Bobrowski K, Schöneich C, Asmus KD (1991) OH-induced oxidation of cyclo-Met-Met. A search for a complexedOH-radical. Radiat Phys Chem 37:473–478

53. Houée-Levin C, Bobrowski K (2013) The use of the methods ofradiolysis to explore the mechanisms of free radical modifications inproteins. J Proteomics. doi:10.1016/j.jprot.2013.02.014

54. Land EJ, PrutzWA (1979) Reaction of azide radicals with amino acidsand proteins. Int J Radiat Biol Relat Stud Phys ChemMed 36(1):75–83

55. DeFelippis MR, Murthy CP, Faraggi M, Klapper MH (1989) Pulseradiolytic measurement of redox potentials: the tyrosine and trypto-phan radicals. Biochemistry 28(11):4847–4853

56. Prutz WA (1990) Free radical transfer involving sulphur peptidefunctions. Plenum, New York

57. Lukacs A, Eker APM, Byrdin M, Brettel K, VosMH (2008) Electronhopping through the 15 A triple tryptophan molecular wire in DNA

photolyase occurs within 30 ps. J Am Chem Soc 130(44):14394–14395

58. Tarboush NA, Jensen LMR, Yukl ET, Geng J, Liu A, Wilmot CM,Davidson VL (2011) Mutagenesis of tryptophan199 suggests thathopping is required for MauG-dependent t ryptophantryptophylquinone biosynthesis. Proc Natl Acad Sci USA 108(41):16956–16961

59. Berges J, de Oliveira P, Fourre I, Houée-Levin C (2012) The one-electron reduction potential of methionine-containing peptides de-pends on the sequence. J Phys Chem B 116(31):9352–9362

60. Stuart-Audette M, Blouquit Y, Faraggi M, Sicard-Roselli C, Houée-Levin C, Jolles P (2003) Re-evaluation of intramolecular long-rangeelectron transfer between tyrosine and tryptophan in lysozymes.Evidence for the participation of other residues. Eur J Biochem270(17):3565–3571

61. Sicard-Roselli C, Lemaire S, Jacquot JP, Favaudon V, Marchand C,Houée-Levin C (2004) Thioredoxin Ch1 of Chlamydomonasreinhardtii displays an unusual resistance toward one-electron oxi-dation. Eur J Biochem 271(17):3481–3487

62. Kadlcik V, Sicard-Roselli C, Mattioli TA, Kodicek M, Houée-Levin C (2004) One-electron oxidation of beta-amyloid peptide:sequence modulation of reactivity. Free Radic Biol Med 37(6):881–891

63. Mozziconacci O, Mirkowski J, Rusconi F, Pernot P, Bobrowski K,Houée-Levin C (2007) Superoxide radical anions protect enkephalinfrom oxidation if the amine group is blocked. Free Radic Biol Med43(2):229–240

64. Kiselar JG, Chance MR (2010) Future directions of structural massspectrometry using hydroxyl radical footprinting. J Mass Spectrom45(12):1373–1382

65. Vahidi S, Stocks BB, Liaghati-Mobarhan Y, Konermann L (2012)Mapping pH-induced protein structural changes under equilibriumconditions by pulsed oxidative labeling and mass spectrometry. AnalChem 84(21):9124–9130

66. Favaudon V, Tourbez H, Houée-Levin C, Lhoste JM (1990) CO2.-radical induced cleavage of disulfide bonds in proteins. A gamma-rayand pulse radiolysis mechanistic investigation. Biochemistry 29(49):10978–10989

67. Arnold K, Bordoli L, Kopp J, Schwede T (2006) The SWISS-MODEL workspace: a web-based environment for protein structurehomology modelling. Bioinformatics 22(2):195–201

68. Schwede T, Kopp J, Guex N, Peitsch MC (2003) SWISS-MODEL:an automated protein homology-modeling server. Nucleic Acids Res31(13):3381–3385

69. Guex N, Peitsch MC (1997) SWISS-MODEL and the swiss-PdbViewer: an environment for comparative protein modeling. Elec-trophoresis 18(15):2714–2723

70. Humphrey W, Dalke A, Schulten K (1996) VMD: visual moleculardynamics. J Mol Graph 14(1):33–8–27–8.

8798 V. Berthelot et al.