Dynamic Article LinksC<Analyst

Cite this: Analyst, 2011, 136, 1739

www.rsc.org/analyst PAPER

Publ

ishe

d on

21

Febr

uary

201

1. D

ownl

oade

d by

Uni

vers

ity o

f Pr

ince

Edw

ard

Isla

nd o

n 31

/10/

2014

06:

48:1

5.

View Article Online / Journal Homepage / Table of Contents for this issue

Induction and identification of disulfide-intact and disulfide-reduced b-subunitof Shiga toxin 2 from Escherichia coli O157:H7 using MALDI-TOF-TOF-MS/MS and top-down proteomics†

Clifton K. Fagerquist* and Omar Sultan

Received 12th November 2010, Accepted 23rd January 2011

DOI: 10.1039/c0an00909a

The disulfide-intact and disulfide-reduced b-subunit of Shiga toxin 2 (b-Stx2) from Escherichia coli

O157:H7 (strain EDL933) has been identified by matrix-assisted laser desorption/ionization time-of-

flight-time-of-flight tandem mass spectrometry (MALDI-TOF-TOF-MS/MS) and top-down

proteomic analysis using software developed in-house. E. coli O157:H7 was induced to express Stx2 by

culturing on solid agar media supplemented with 10–50 ng mL�1 of ciprofloxacin (CP). Bacterial cell

lysates at each CP concentration were analyzed by MALDI-TOF-MS. A prominent ion at mass-to-

charge (m/z)�7820 was observed for the CP concentration range: 10–50 ng mL�1, reaching a maximum

signal intensity at 20 ng mL�1. Complex MS/MS data were obtained of the ion at m/z �7820 by post-

source decay resulting in top-down proteomic identification as the mature, signal peptide-removed,

disulfide-intact b-Stx2. Eight fragment ion triplets (each spaced Dm/z �33 apart) were also observed

resulting from backbone cleavage between the two cysteine residues (that form the intra-molecular

disulfide bond) and symmetric and asymmetric cleavage of the disulfide bond. The middle fragment ion

of each triplet, from symmetric disulfide bond cleavage, was matched to an in silico fragment ion

formed from cleavage of the backbone at a site adjacent to an aspartic acid or glutamic acid residue.

The flanking fragment ions of each triplet, from asymmetric disulfide bond cleavage, were not matched

because their corresponding in silico fragment ions are not represented in the database. Easier to

interpret MS/MS data were obtained for the disulfide-reduced b-Stx2 which resulted in an improved

top-down identification.

Introduction

Shiga toxins (Stxs) are one of the causative agents of severe

foodborne illness due to Escherichia coli O157:H7 and other

Shiga toxin-containing E. coli (STEC). Stx infection can result in

hemolytic uremic syndrome (HUS), kidney failure and even

death.1,2 Stx is an AB5 toxin comprised of an a-subunit and five

identical b-subunits. After translation, the a- and b-subunits are

shuttled to the bacterial periplasm where, after their N-terminal

signal peptides have been removed, they self-assemble into the

holotoxin. The five b-subunits form a non-covalent donut-sha-

ped complex. The a-subunit is positioned primarily on one side

of the donut but also extends into the central pore. On the side

opposite to the a-subunit attachment, the b-pentamer forms

a highly specific non-covalent attachment to the surface of the

human intestine via glycosphingolipid globotriaosylceramide

(Gb3) receptors on the epithelial cell surface. In a complicated

Western Regional Research Center, Agricultural Research Service, U.S.Department of Agriculture, 800 Buchanan Street, Albany, CA, 94710,USA. E-mail: clifton.fagerquist@ars.usda.gov

† Electronic supplementary information (ESI) available: See DOI:10.1039/c0an00909a

This journal is ª The Royal Society of Chemistry 2011

series of steps, the holotoxin is transported into the eukaryotic

cell where the a-subunit eventually undergoes proteolytic

cleavage by a human enzyme. This truncated a-subunit disrupts

ribosomal protein biosynthesis in the cytosol contributing to cell

damage and death.3

Two of the most important Shiga toxin genes are stx1 and

stx2. Stx2 is more often linked to severe foodborne illness. There

are a number of Stx1 and Stx2 variants which have slightly

different amino acid sequences in the two subunits. A number of

bacterial microorganisms may carry the stx gene, however

E. coli, and specifically the O157:H7 serotype, is most often

linked to foodborne illness that involves the presence of Stx. A

number of other non-O157:H7 serotypes of E. coli (e.g. O26,

O45, O103, O111, O121, O145, etc.) may also carry genes for stx1

and/or stx2.4–8 Unlike O157:H7, these non-O157 serotypes are

not currently considered, from a regulatory standpoint, as an

adulterant although they may cause illness as severe as that

caused by the O157:H7 serotype.

Typically, polymerase chain reaction (PCR) is used to deter-

mine if a microorganism carries stx genes by amplifying short

segments of the gene(s) with specific primers.6,8 Expression of Stx

toxin is often determined using an enzyme immunoassay8,9 or

Analyst, 2011, 136, 1739–1746 | 1739

Publ

ishe

d on

21

Febr

uary

201

1. D

ownl

oade

d by

Uni

vers

ity o

f Pr

ince

Edw

ard

Isla

nd o

n 31

/10/

2014

06:

48:1

5.

View Article Online

a Vero cell toxicity assay.10 Use of mass spectrometry-based

techniques for detection, identification or analysis of Stx has

been limited. Williams et al. reported detection of the b-subunit

of Stx1 from bacterial cell lysates in one of two closely-related

E. coli O157:H7 strains using liquid chromatography/electro-

spray ionization mass spectrometry (LC/ESI-MS).11 Kitova et al.

used ESI and high resolution Fourier transform ion cyclotron

resonance mass spectrometry (FT-ICR-MS) to examine the

intact AB5 complex of Stx1 and Stx2. The holotoxin self-

assembled in solution due to hydrophobic and/or non-covalent

forces, and the quaternary structure appeared to be retained into

the gas phase. Temperature and pH differences in solution were

reflected by variations in ESI charge state distributions in the gas

phase.12

Increasingly, matrix-assisted laser desorption/ionization

(MALDI) time-of-flight mass spectrometry (MALDI-TOF-MS)

and pattern recognition analysis are being used in the ‘‘finger-

printing’’ of microorganisms.13–17 MALDI-TOF-TOF tandem

mass spectrometry (MS/MS) has been used to isolate, fragment,

and identify intact protein ions from microorganisms in

conjunction with top-down proteomic analysis.18–20 Very

recently, our group reported top-down proteomic identification

of protein biomarkers from bacterial cell lysates of pathogenic

and non-pathogenic strains of E. coli cultured using conventional

solid agar media.19 Proteins were ionized by MALDI and

analyzed by TOF-TOF-MS/MS. Web-based software, developed

in-house, was used to rapidly compare the m/z of MS/MS frag-

ment ions to a database of in silico fragment ions derived from

bacterial protein sequences.20,21 In the current report, we

demonstrate definitive identification of the b-subunit of Stx2

(b-Stx2) from bacterial cell lysates of E. coli O157:H7 grown on

solid agar media supplemented with an antibiotic that is

a powerful inducer of the bacterial SOS-response.22 The

expressed b-Stx2 was ionized, analyzed and identified by

MALDI-TOF-TOF-MS/MS and top-down proteomics using

our in-house software. Fragmentation of the singly charged

b-Stx2 ion resulted in unusually complex MS/MS data due to the

presence of an intra-molecular disulfide bond. Analysis of the

disulfide-reduced b-Stx2 resulted in less complex MS/MS data

and an improved top-down identification.

Portions of this work were presented at the 239th National

American Chemical Society meeting (Spring 2010, San Francisco,

CA).23

Materials and methods

E. coli O157:H7 and induction of Stx2

Safety considerations: E. coli O157:H7 is a class II bacterial

microorganism. Microbiological manipulations were conducted

in a class II biohazard cabinet.

Approximately 1 mL of frozen stock E. coli O157:H7 (strain

ATCC # 43895, EDL933) was transferred to a 50 mL sterile

conical tube containing 25 mL of Luria-Bertani (LB) broth

(Becton Dickinson, Franklin Lakes, NJ) and incubated over-

night at 37.5 �C under static conditions. A 100 mL aliquot from

the overnight suspension was transferred to LB agar plates

(Becton Dickinson, Franklin Lakes, NJ) supplemented with 10,

20, 30, 40, or 50 ng mL�1 of the antibiotic ciprofloxacin (CP)

1740 | Analyst, 2011, 136, 1739–1746

(Fluka BioChemica, Buchs, Switzerland) to induce Stx2 expres-

sion. As a negative control, one plate was streaked in the absence

of CP (0 ng mL�1). The plates were incubated overnight at 37.5�C. Bacterial cells from each of the six plates were harvested

using a 1 mL sterile transfer loop, and transferred to a separate

2 mL microcentrifuge tube containing 300 mL of a solution of

67% HPLC-grade water (Honeywell, Burdick and Jackson,

Muskegon, WI), 33% HPLC-grade acetonitrile (Acros Organics,

Geel, Belgium) and 0.2% trifluoroacetic acid (Sigma-Aldrich, St.

Louis, MO) and approximately 100 mg of 0.1 mm zirconia–silica

beads (BioSpec Products, Bartlesville, OK). The tubes were

capped and homogenized by bead-beating for 60 seconds on

a reciprocating shaker (Mini-Beadbeater; Biospec Products,

Bartlesville, OK) after which they were centrifuged at 14 000 rpm

for 2 minutes. The supernatant was analyzed by mass spec-

trometry and tandem mass spectrometry.

Disulfide reduction of b-Stx2

Reduction of the single disulfide bond of b-Stx was performed

using the following protocol. Bacterial cells were grown over-

night on LB agar plates containing 20 ng mL�1 CP and harvested

as described previously. However, the extraction solution con-

sisted of 300 mL of HPLC-grade water. Twenty mL of bacterial

cell supernatant was mixed with 1 mL of 1 M molecular grade

dithiothreitol (DTT, Sigma-Aldrich, St. Louis, MO) followed by

incubation at 70 �C for 10 minutes in order to reduce the intra-

molecular disulfide bond. The sample was then analyzed by mass

spectrometry and tandem mass spectrometry.

Mass spectrometry and tandem mass spectrometry

A 1 mL aliquot of bacterial cell lysate supernatant was spotted

onto a 384-well stainless steel target plate and allowed to dry at

room temperature. A 1 mL aliquot of saturated MALDI matrix

solution was spotted on top of the dried sample spot and allowed

to dry at room temperature. Matrix solutions consisted of

a saturated solution of either 3,4-dihydroxy-cinnamic acid

(caffeic acid), a-cyano-4-hydroxy-cinnamic acid (HCCA), or

3,5-dimethoxy-4-hydroxy-cinnamic acid (sinapinic acid) in 67%

HPLC-grade water, 33% HPLC-grade acetonitrile, and 0.2%

trifluoroacetic acid. All MALDI matrices were obtained from

commercial suppliers (Fluka Analytical, Switzerland) at a purity

of $99.0% and were used without further recrystallization.

Standards for calibrating in MS linear mode were cytochrome C

from horse heart (ICN Biochemicals, Cleveland, OH), lysozyme

from chicken egg white (Sigma-Aldrich, St. Louis, MO) and

myoglobin from horse skeletal muscle (Sigma-Aldrich, St. Louis,

MO). [Glu1]-fibrinopeptide B human (Sigma-Aldrich, St. Louis,

MO) was used to calibrate in MS/MS reflectron-mode.

MS and MS/MS analyses were performed using a 4800 Plus

MALDI-TOF/TOF mass spectrometer (Applied Biosystems,

Foster City, CA) equipped with a pulsed solid-state YAG laser

(200 Hz, 355 nm, 5 ns pulse width). MS analysis was conducted in

linear positive ion mode using default settings for a mid-mass

analysis at an acceleration voltage of 20 kV. The linear TOF was

externally calibrated using a protein standard mixture of

cytochrome C (Ave. MW ¼ 12 361 Da), lysozyme (Ave.

This journal is ª The Royal Society of Chemistry 2011

Publ

ishe

d on

21

Febr

uary

201

1. D

ownl

oade

d by

Uni

vers

ity o

f Pr

ince

Edw

ard

Isla

nd o

n 31

/10/

2014

06:

48:1

5.

View Article Online

MW¼ 14 306 Da), and myoglobin (Ave. MW¼ 16 953 Da). MS

data were collected using 550–1000 laser shots.

For MS/MS analysis, the instrument was operated using the

default settings of the MS/MS 1 kV (or 2 kV) positive ion

sensitivity mode. The metastable suppressor was ‘‘on’’. The

timed-ion selector (TIS) was operated at a resolution of �100

Da. No collision gas was introduced into the collision cell, and

precursor ion fragmentation was the result of post-source decay

(PSD) using higher than normal laser fluence.18,20 The reflectron-

TOF was externally calibrated using fragment ions of MS/MS of

the +1 charge state of [Glu1]-fibrinopeptide B (MW 1570.60 Da):

m/z 175.119, 684.346, 813.389, 1056.475, and 1441.634 (mono-

isotopic). Positive ions were accelerated from the first source at

8.0 kV, de-accelerated to 1 kV (or 2 kV) in the collision cell, and

then re-accelerated to 15 kV in the second source. MS/MS data

were collected using 10 000 to 20 000 laser shots.

All data were processed using Data Explorer (Version 4.9,

Applied Biosystems, Foster City, CA). Raw MS data were pro-

cessed using a noise filter (correlation factor¼ 0.7). Raw MS/MS

data were processed in the following sequence. First, an

advanced baseline correction (peak width ¼ 32, flexibility ¼ 0.5,

degree ¼ 0.1) is applied, followed by noise removal (standard

deviation ¼ 2) and finally a Gaussian smooth (filter width ¼ 31

points). Processed MS and MS/MS data were centroided,

exported as an ASCII file (m/z versus absolute intensity) and

uploaded to their respective databases in the USDA software.

Top-down proteomic analysis

The web-based software developed in our laboratory for top-

down proteomic analysis using MALDI-TOF-TOF-MS/MS has

been described in detail elsewhere.20,21 Briefly, the m/z of MS/MS

fragment ions are compared to the m/z of in silico fragment ions

derived from bacterial protein sequences that have the same

MW as the protein biomarker ion. Bacterial protein sequences

were downloaded from the ExPASy website (www.expasy.org)

using their TagIdent software. TagIdent criteria for protein

retrieval were as follows: pI range: 0.00–14.00, protein MW

(Da) ¼ (m/z �1) protein biomarker; protein MW tolerance �10

Da, organism classification: bacteria, cysteines in their oxidized

form (however, no difference in sequence retrieval was observed

when the same search was performed with cysteines in their

reduced form). No sequence tag was utilized for the retrieval.

Databases: UniProtKB/Swiss-Prot (release 57.15; 02-Mar-10)

and UniProtKB/TrEMBL (release 40.15; 02-Mar-10). On the

possibility that the protein biomarker may be modified with the

most common PTM among bacterial proteins, i.e. N-terminal

methionine cleavage, a second retrieval was performed with the

protein MW ¼ [(m/z �1) + 131] protein biomarker. Two multi-

protein sequence FASTA files were downloaded from ExPASy.

The in silico bacterial protein sequences in these files were batch

processed using a beta-version of GPMAW software (8.01a5) to

generate individual protein sequence files containing the

ExPASy accession designation, the protein name, bacteria

classification, the protein sequence, the calculated average MW

of the sequence and the in silico fragment ions identified by the

their average m/z, type of ion (a, b, b � 18, y, y � 17, y � 18) and

the two amino acid residues immediately adjacent to the site of

polypeptide backbone cleavage that resulted in the fragment ion.

This journal is ª The Royal Society of Chemistry 2011

The two multi-sequence FASTA files were first processed with

cysteines in their oxidized form and then processed with cyste-

ines in their reduced form. Protein sequences with signal

peptides were identified from their larger than normal file sizes

and processed as previously described.20 GPMAW-generated

files were uploaded to their respective database in the USDA

software.

The USDA software uses a simple peak matching scoring

algorithm20,21 and Demirev’s p-value algorithm18 to indepen-

dently score/rank MS/MS-to-in silico identifications. Analyses

were performed using the following comparison parameters:

minimum MS/MS fragment ion intensity threshold of 1% to 2%;

fragment ion m/z tolerance of �1.0; protein molecular weight

(MW) tolerance of �10 Da. MS/MS-to-in silico comparisons

were performed using both a non-residue-specific in silico as well

as a D-, E-, P-specific comparison.

Results and discussion

Fig. S-1 (ESI†) shows images of the bacterial growth from

overnight incubation. CP concentrations in the solid agar media

are indicated. Based on simple visual inspection, maximum

bacterial growth occurred when no antibiotic is present. As the

concentration of antibiotic was increased, a general clearing of

the plate was observed concomitant with the appearance of

discrete isolated colonies. At the highest CP concentration,

bacterial growth appears to be strongly inhibited, and the

number of discrete colonies is reduced to four (indicated by

yellow arrows).

Adjacent to each plate image in Fig. S-1† is the corresponding

MALDI-TOF-MS spectra of the bacterial cell lysate obtained

for that plate. In the absence of an antibiotic (0 ng mL�1),

a number of protein ions are observed (also shown in the top

spectrum of Fig. 1). Peaks at m/z 7280, 7716, 8334, 9748, 10 485

were previously identified by MALDI-TOF-TOF-MS/MS and

top-down proteomics as the cold shock protein C, YahO, YjbJ,

acid-stress chaperone-like protein: HdeA, and the homeobox

(YbgS) protein, respectively.19 The MS spectrum at the 10 ng

mL�1 CP concentration shows significant change in the relative

intensity of protein ions as well as the sudden appearance of

a peak at m/z 7819 which is absent when no antibiotic is present

(top spectrum). At the 20 ng mL�1 CP concentration, the peak

at m/z �7819 becomes the dominant peak in the spectrum

(also shown in the bottom spectrum of Fig. 1). It is possible that

the relative abundance of the protein suppresses ionization of

other proteins in the sample. At 30, 40 and 50 ng mL�1 CP

concentrations, the absolute intensity of the peak at m/z�7820 is

observed to decline with a concomitant increase in the intensities

of other protein ion peaks. At 50 ng mL�1 CP concentration,

bacterial growth was inhibited to such an extent that there were

insufficient cells to completely fill a 1 mL loop which resulted in

weak ion signal.

The peak at m/z �7819 (20 ng mL�1 of CP) was analyzed by

MS/MS (Fig. 2), and the m/z of the fragment ions generated were

compared against a database of in silico fragment ions derived

from bacterial protein sequences having the same MW as the

protein biomarker. The results of a non-residue-specific

comparison are shown in Table S-1-A (ESI†). The top identifi-

cation of both scoring algorithms was the sequence of the mature

Analyst, 2011, 136, 1739–1746 | 1741

Fig. 1 MALDI-TOF-MS data (using caffeic acid) of cell lysates of E. coli O157:H7 harvested from LB agar plates with no ciprofloxacin (CP) (top

spectrum) and LB supplemented with 20 ng mL�1 of CP (bottom spectrum).

Publ

ishe

d on

21

Febr

uary

201

1. D

ownl

oade

d by

Uni

vers

ity o

f Pr

ince

Edw

ard

Isla

nd o

n 31

/10/

2014

06:

48:1

5.

View Article Online

b-Stx2 with the 19-residue N-terminal signal peptide removed

and the intra-molecular disulfide bond intact.3 The in silico

sequence of the disulfide-reduced b-Stx2 was not ranked within

the top 30 identifications. As shown, the sequence of the top

identification is fully homologous between six E. coli O157:H7

strains (including the EDL933 strain used in the present study),

one E. coli O157:H� strain, one E. coli O111 strain, and one

strain of Acinetobacter haemolyticus.24 The average MW of this

sequence of b-Stx2 is 7815.62 Da which incorporates removal of

a 19-residue N-terminal signal peptide and formation of one

intra-molecular disulfide bond. Another top-ranked b-Stx2

sequence (in silico ID: 43943) from an E. coli O157 strain is

slightly different from the other b-Stx2 sequence due to amino

acid substitutions at residue 5 (Q 4 K) and at residue 51 (I 4 L)

resulting in an average MW ¼ 7815.58 Da.25 Unfortunately, the

residue variation between these two b-Stx2 sequences results in

nearly identical MWs and do not involve a substitution that

might result in a distinguishing sequence-specific fragmentation,

such as with aspartic acid, glutamic acid or proline residues.19 In

consequence, it is difficult to discriminate between these two

sequences even by MS/MS.

Non-digested, singly charged protein ions do not routinely

dissociate to provide fragment ions corresponding to a series of

adjacent amino acid residues (i.e. a sequence ‘‘tag’’). In conse-

quence, de novo sequencing is challenging at best. This is

primarily due to the size of a protein ion and the lack of protons

available to assist in polypeptide backbone fragmentation. In

1742 | Analyst, 2011, 136, 1739–1746

consequence, a singly charged protein ion often will fragment at

aspartic, glutamic and proline residues which are randomly

distributed throughout a sequence. In the current study, only

a certain number of matched MS/MS fragment ions were found

to correspond to polypeptide backbone cleavage at adjacent

amino acid residues (at most three) as shown in Fig. 2, 3 and 5

and S-2†. A BLAST search of these residue trimers failed due to

the shortness of the sequence ‘‘tags’’.

In order to more accurately measure the MW of the b-Stx2, the

bacterial cell lysate at 20 ng mL�1 of CP was analyzed by nano-

ESI using an externally calibrated quadrupole-TOF-MS.

Although the MS data were congested due to charge state

envelopes from multiple proteins (data not shown), we detected

the +4, +5, +6, +7 charge states of b-Stx2 resulting in an average

MW of 7815.4 � 0.1 Da. We also analyzed the bacterial cell

lysate at 20 ng mL�1 of CP by 1-D SDS-PAGE and bottom-up

proteomics. A prominent gel band at �8 kDa was identified as

b-Stx2 (Fig. S-3 and Table S-4-A, S-4-B, S-5-A and S-5-B, ESI†).

A MS/MS-to-in silico comparison was also performed using

a D-, E-, P-specific comparison in Table S-1-B (ESI†). Once

again, the top identification and the 1st runner-up identification

are b-Stx2 sequences. The top/runner-up score differential is

wider with this residue-specific comparison in contrast to the

top/runner-up score differential for the non-residue-specific

comparison in Table S-1-A†. In addition, the runner-up identi-

fications and/or rankings are different for the D-, E-, P-specific

comparison in contrast to the non-residue-specific comparison

This journal is ª The Royal Society of Chemistry 2011

Fig. 2 MS/MS of the precursor ion at m/z �7819 shown in Fig. 1 (at the 20 ng mL�1 CP concentration). Fragment ions matched to in silico fragment

ions of the b-Stx2 sequence are indicated by the ion type/number as well as the two amino acid residues immediately adjacent to the site of backbone

cleavage. A question mark indicates that either the MS/MS fragment ion was below the 2% threshold or the fragment ion error exceeded the �m/z 1.0

tolerance used in the comparison. Fragment ions highlighted with a star may be internal fragment ions from double cleavages of the backbone. Insets

display expanded mass ranges of fragment ion triplets generated by backbone cleavage between two cysteine residues involved in an intra-molecular

disulfide bond with symmetric and asymmetric cleavage of the disulfide bond. The amino acid sequence of b-Stx2 show sites of polypeptide cleavage (*)

that correspond to the fragment ions observed.

Publ

ishe

d on

21

Febr

uary

201

1. D

ownl

oade

d by

Uni

vers

ity o

f Pr

ince

Edw

ard

Isla

nd o

n 31

/10/

2014

06:

48:1

5.

View Article Online

suggesting that the non-b-Stx2 identifications in both compari-

sons are the result of random false matches.

As shown in Fig. 2, MS/MS fragment ions matched to in silico

fragment ions of the b-Stx2 sequence are identified by their m/z,

ion type/number and the two residues immediately adjacent to

the site of polypeptide cleavage. As can be seen many of the

fragment ion matches occur from cleavage of the polypeptide

backbone adjacent to D or E residues. This has been previously

explained as due to the transfer of a proton from the side-chain of

a D or E residue to the polypeptide backbone facilitating its

fragmentation adjacent to these specific residues.26 A number of

other prominent fragment ions in Fig. 2 did not have a match to

Fig. 3 Amino acid sequence of disulfide-intact b-Stx2 sequence of E. coli

O157:H7 (EDL933). A 19-residue N-terminal signal peptide is shown in

outline. Two cysteine residues (boxed) are linked by an intra-molecular

disulfide bond (S/S). An asterisk indicates a site of backbone cleavage

on the basis of matched fragment ions in Fig. 2. A question mark indi-

cates either that the MS/MS fragment ion was below the 2% threshold or

the fragment ion error exceeded the �m/z 1.0 tolerance used in the

comparison.

This journal is ª The Royal Society of Chemistry 2011

in silico fragment ions of the b-Stx2 sequence. These fragment

ions (shown in insets in Fig. 2) appeared at a m/z that was �33

higher (or �33 lower) than the m/z of a matched fragment ion

peak. The appearance of a triplet of fragment ion peaks sepa-

rated by �m/z 33 suggests they are the result of symmetric and

asymmetric cleavage of an intra-molecular disulfide bond.

Fig. 3 shows the amino acid sequence of b-Stx2 with the

backbone cleavage sites symbolized with an asterisk. Above (or

below) each cleavage site is the corresponding in silico fragment

ion match. Matched fragment ions that have flanking fragment

ions at m/z��33 are the result of backbone cleavage between the

two cysteine residues involved in the putative disulfide bond.

They are: y43, y46, y46 � 17, y53, y53 � 17, y54, y54 � 18 and y55.

Matched fragment ions that are the result of backbone cleavage

not between the two cysteine residues do not have flanking

fragment ions at m/z ��33. They are: b56, b57, b57 � 18, b58,

b58 � 18, b59, b61, b63, b64, b64 � 18, y11 � 17, y68 � 17, y69 � 18.

This can be explained as follows. If polypeptide backbone frag-

mentation occurs between two cysteine residues that are involved

in an intra-molecular disulfide bond, then the resulting two

peptides are still linked by the disulfide bond. The disulfide bond

quickly ruptures by symmetric and asymmetric cleavage giving

the characteristic triplet of fragment ion peaks. If backbone

cleavage occurs, but at a site that is not between the two cysteine

residues, then the fragment ion formed contains both cysteine

residues and the putative disulfide bond. The detected m/z of the

Analyst, 2011, 136, 1739–1746 | 1743

Fig. 4 Possible internal fragment ion from double cleavage of the

backbone by a combination of b-type and y-type cleavages (outside the

disulfide loop). The MS/MS fragment ion at m/z 6257.2 (Fig. 2) may be

formed from y68 and b57 cleavages resulting in an N-terminal amino

group, a C-terminal acylium ion (–C^O+) and the side-chain of the C-

terminal glutamic acid with a carboxylate anion functional group

(–COO�). Dissociative loss of H2 via an elimination reaction in the side-

chain of glutamic acid would result in a m/z close to the observed value.

Fig. 5 Amino acid sequence of disulfide-reduced b-Stx2 sequence of

E. coli O157:H7 (EDL933). A 19-residue N-terminal signal peptide is

shown in outline. Two cysteine residues (boxed) are displayed in their

reduced form. An asterisk indicates a site of backbone cleavage on the

basis of matched fragment ions in Fig. S-2 (ESI†).

Publ

ishe

d on

21

Febr

uary

201

1. D

ownl

oade

d by

Uni

vers

ity o

f Pr

ince

Edw

ard

Isla

nd o

n 31

/10/

2014

06:

48:1

5.

View Article Online

fragment ion would be the same if the disulfide bond were intact

or ruptured. It is possible that the putative disulfide bond

ruptures before polypeptide backbone cleavage. However, the

appearance of the Dm/z ��33 fragment ion triplet would still be

determined by whether the site of polypeptide backbone cleavage

is between the two formerly disulfide-linked cysteine residues

(or not). Previous researchers have reported fragmentation of

singly27 and doubly and triply28 protonated peptide ions that

have a single intra-molecular disulfide bond. It has been sug-

gested that a salt-bridge mechanism,29 facilitated by non-mobile

protons, may be responsible for disulfide bond cleavage observed

in peptides ionized by MALDI.27,30,31

The in silico fragment ions generated by GPMAW and

uploaded to our in silico database assume that any disulfide

bonds present in a protein sequence cleave symmetrically. In

consequence, MS/MS fragment ions that are the result of disul-

fide bonds cleaving asymmetrically will not be correctly matched

to their corresponding in silico fragment ions and may be

incorrectly matched to other in silico fragment ions. In order to

determine what effect (if any) these asymmetrically disulfide

cleaved peaks had on scoring/ranking, we removed these 16

fragment ion peaks and analyzed the resulting MS/MS data

using a non-residue-specific MS/MS-to-in silico comparison and

are shown in Table S-2-A (ESI†). Once again, the top identifi-

cation was b-Stx2 and there was no significant change in the

score differential between the top and lower ranking identifica-

tions compared to the analysis when these 16 peaks were

included. A MS/MS-to-in silico comparison was also performed

using a D-, E-, P-specific comparison, and a similar result was

obtained as shown in Table S-2-B (ESI†).

Five prominent fragment ions in Fig. 2 at m/z 6257, 6893, 6911,

7009 and 7026 (highlighted with a star) were also not matched to

in silico fragment ions of b-Stx2 sequence. These fragment ions

may be internal fragment ions resulting from double cleavage of

the polypeptide backbone. For example, the fragment ion at m/z

6257 may be formed from a combination of y68 and b57 cleavages

resulting in an N-terminal amino group and a C-terminal

acylium ion (NH2/C^O+). If the C-terminal glutamic acid of

this fragment transferred a proton to the backbone facilitating its

fragmentation, its side-chain would then have a carboxylate

anion (–COO�). The carboxylate anion may then react with the

acylium cation. Finally, an elimination reaction in the glutamic

acid side-chain could result in dissociative loss of H2 (Fig. 4). A

similar rearrangement mechanism may explain the fragment ions

at m/z 6893 (from y68 and b64 cleavages) and possibly m/z 7009

(from y69 and b64 cleavages). However, it is not clear what

mechanism or rearrangement may be responsible for the un-

matched fragment ions at m/z 6911 and 7026. A c-type cleavage

may be involved, however no single cleavage c-type fragment

ions were observed in the spectrum. We have previously reported

possible evidence for internal fragment ions from double

cleavage of the polypeptide backbone of singly charged

(protonated) protein ions by MALDI-TOF-TOF-MS/MS,

however this is an area that requires further investigation.32,33

Fig. S-2 (ESI†) shows MS/MS analysis of the putative disul-

fide-reduced b-Stx2 at m/z �7820 from a DTT-treated bacterial

cell lysate cultured on solid agar supplemented with 20 ng mL�1

of CP. The prominent fragment ion triplets, present in the

MS/MS data of the disulfide-intact b-Stx2 (Fig. 2), are either

1744 | Analyst, 2011, 136, 1739–1746

absent or significantly reduced in abundance as shown in Fig.

S-2†. The un-matched fragment ions in Fig. 2 that may be the

result of double cleavages of the backbone are also absent in

Fig. S-2† suggesting that conversion of b-Stx2 to a linear

polypeptide chain without the large disulfide loop appears to

suppress formation of these fragment ions. A non-residue-

specific MS/MS-to-in silico comparison of this spectrum is shown

in Table S-3-A (ESI†). The top identification (with 30 matched

MS/MS fragment ions) is the disulfide-reduced b-Stx2 sequence.

The 1st runner-up identification (with five fewer matched

MS/MS fragment ions) is the disulfide-intact b-Stx2 sequence.

The next lower ranked identification (an uncharacterized protein

from a strain of Methylobacillus flagellatus) has only 17 matched

MS/MS fragment ions. A MS/MS-to-in silico comparison was

also performed using a D-, E-, P-specific comparison (Table

S-3-B, ESI†). Once again, the top identification is disulfide-

reduced b-Stx2 with the disulfide-intact b-Stx2 as the 1st

runner-up. However, the next lower ranked identification is an

uncharacterized protein from a strain of Rhodobacter aeroides

(which is different from that of the non-residue-specific

comparison). In contrast with top-down identification of

disulfide-intact b-Stx2, top-down identification of disulfide-

reduced b-Stx2 resulted in a lower P-value (a more significant

identification) and a higher USDA score as well as a wider score

differential between b-Stx2 and non-b-Stx2 identifications.

Although the number of MS/MS fragment ions for the disulfide-

reduced b-Stx2 was approximately 2/3 that of the disulfide-intact

b-Stx2, the MS/MS data of the disulfide-reduced b-Stx2 have

a flatter baseline and better S/N over the entire mass range which

resulted in an improved identification.

Fig. 5 shows the backbone cleavage sites (symbolized with

an asterisk) of the disulfide-reduced b-Stx2 sequence. Above

This journal is ª The Royal Society of Chemistry 2011

Publ

ishe

d on

21

Febr

uary

201

1. D

ownl

oade

d by

Uni

vers

ity o

f Pr

ince

Edw

ard

Isla

nd o

n 31

/10/

2014

06:

48:1

5.

View Article Online

(or below) each cleavage site is the corresponding in silico

fragment ion match(es). Interestingly, there are 15 cleavage sites

between the two cysteine residues for the disulfide-reduced

b-Stx2, whereas there are only six in the disulfide-intact b-Stx2

(Fig. 3), and there are three backbone cleavage sites not between

the two cysteine for the disulfide-reduced b-Stx2 but there are

eight for the disulfide-intact b-Stx2 (Fig. 3). Clearly, reduction of

the intra-molecular disulfide bond has altered the number and, to

some extent, the location of the backbone cleavage sites.

Presumably, differences in energy re-distribution for a linear

chain versus a large looped structure may have a significant effect

on the fragmentation differences observed. Alternatively, the

energy expended to cleave the disulfide bond (either symmetri-

cally or asymmetrically) in the gas phase for the disulfide-intact

b-Stx2 may reduce the energy available for fragmentation of the

backbone. The simpler linear chain structure of the disulfide-

reduced b-Stx2 may improve top-down identification by

reducing (or eliminating) fragment ions from more complex

dissociation mechanisms (e.g. internal fragment ions, etc.) that

are not currently represented in our in silico database.

Conclusions

We have identified the disulfide-intact and disulfide-reduced

b-subunit of Stx2 from the bacterial cell lysate of E. coli O157:H7

(strain EDL933) using MALDI-TOF-TOF-MS/MS and top-

down proteomic techniques. Expression of Stx2 was induced by

culturing E. coli O157:H7 on solid agar media supplemented with

CP. Maximum signal intensity of b-Stx2 occurred at a concen-

tration of 20 ng mL�1 of CP. PSD of the singly charged disulfide-

intact b-Stx2 ion resulted in complex MS/MS data suggesting the

presence of a single intra-molecular disulfide bond between the

two cysteine residues of this protein sequence consistent with

previous research on this toxin subunit. Backbone cleavage

between the two cysteine residues resulted in a triplet of fragment

ions spaced Dm/z �33 apart suggesting symmetric and asym-

metric cleavage of the disulfide bond. The middle fragment ion of

each fragment ion triplet was correctly matched to its corre-

sponding in silico fragment ion by top-down analysis because

only in silico fragment ions with symmetrically cleaved disulfide

bonds are represented in the in silico database. MS/MS fragment

ions resulting from asymmetrically cleaved disulfide bonds were

not matched to in silico fragment ions. As expected, cleavage of

the backbone at a site not between the two cysteine residues did

not result in a triplet of fragment ions. Five prominent fragment

ions, not part of a fragment ion triplet and thus not the result of

backbone fragmentation between the two cysteine residues, may

be internal fragment ions formed as a result of a double cleavage

of the backbone outside the putative disulfide loop. MS/MS of

the disulfide-reduced b-Stx2 ion resulted in simpler, easier to

interpret data that resulted in an improved top-down identifi-

cation compared to the disulfide-intact identification. Inclusion

of a simple disulfide-reduction step in the sample protocol would

appear to benefit top-down identification of disulfide-containing

proteins from bacterial cell lysates. However, top-down analysis

of bacterial cell lysates without disulfide reduction has the

advantage of allowing detection of fragment ion triplets that are

characteristic of the presence of disulfide bond(s) in the mature,

post-translationally modified protein being analyzed by MS/MS.

This journal is ª The Royal Society of Chemistry 2011

Disclaimer

Mention of a brand or firm name does not constitute an

endorsement by the U.S. Department of Agriculture over other

of a similar nature not mentioned. This article is a US Govern-

ment work and is in the public domain in the USA.

Acknowledgements

We wish to thank Anne H. Bates for microbiological assistance

and Dr Robert E. Mandrell for suggesting analysis of Shiga

toxin. We also wish to thank Brandon R. Garbus for useful

discussions.

References

1 H. Uchida, N. Kiyokawa, H. Horie, J. Fujimoto and T. Takeda, Thedetection of Shiga toxins in the kidney of a patient with hemolyticuremic syndrome, Pediatr. Res., 1999, 45(1), 133–137.

2 H. Uchida, N. Kiyokawa, T. Taguchi, H. Horie, J. Fujimoto andT. Takeda, Shiga toxins induce apoptosis in pulmonary epithelium-derived cells, J. Infect. Dis., 1999, 180(6), 1902–1911.

3 L. Johannes and W. Romer, Shiga toxins—from cell biology tobiomedical applications, Nat. Rev. Microbiol., 2010, 8(2), 105–116.

4 T. Slanec, A. Fruth, K. Creuzburg and H. Schmidt, Molecularanalysis of virulence profiles and Shiga toxin genes in food-borneShiga toxin-producing Escherichia coli, Appl. Environ. Microbiol.,2009, 75(19), 6187–6197.

5 Y. Ogura, T. Ooka and A. Iguchi, et al., Comparative genomics revealthe mechanism of the parallel evolution of O157 and non-O157enterohemorrhagic Escherichia coli, Proc. Natl. Acad. Sci. U. S. A.,2009, 106(42), 17939–17944.

6 E. B. Hedican, C. Medus and J. M. Besser, et al. Characteristics ofO157 versus non-O157 Shiga toxin-producing Escherichia coliinfections in Minnesota, 2000–2006, Clin. Infect. Dis., 2009, 49(3),358–364.

7 T. Tasara, M. Bielaszewska, S. Nitzsche, H. Karch, C. Zweifel andR. Stephan, Activatable Shiga toxin 2d (Stx2d) in STEC strainsisolated from cattle and sheep at slaughter, Vet. Microbiol., 2008,131(1–2), 199–204.

8 T. E. Grys, L. M. Sloan, J. E. Rosenblatt and R. Patel, Rapid andsensitive detection of Shiga toxin-producing Escherichia coli fromnonenriched stool specimens by real-time PCR in comparison toenzyme immunoassay and culture, J. Clin. Microbiol., 2009, 47(7),2008–2012.

9 P. J. Gavin, L. R. Peterson and A. C. Pasquariello, et al. Evaluation ofperformance and potential clinical impact of ProSpecT Shiga toxinEscherichia coli microplate assay for detection of Shiga toxin-producing E. coli in stool samples, J. Clin. Microbiol., 2004, 42(4),1652–1656.

10 B. Quinones, S. Massey, M. Friedman, M. S. Swimley and K. Teter,Novel cell-based method to detect Shiga toxin 2 from Escherichia coliO157:H7 and inhibitors of toxin activity, Appl. Environ. Microbiol.,2009, 75(5), 1410–1416.

11 T. L. Williams, S. R. Monday, P. C. Feng and S. M. Musser,Identifying new PCR targets for pathogenic bacteria using top-down LC/MS protein discovery, J. Biomol. Tech., 2005, 16(2), 134–142.

12 E. N. Kitova, G. L. Mulvey and T. Dingle, et al. Assembly andstability of the Shiga toxins investigated by electrospray ionizationmass spectrometry, Biochemistry, 2009, 48(23), 5365–5374.

13 T. Krishnamurthy, P. L. Ross and U. Rajamani, Detection ofpathogenic and non-pathogenic bacteria by matrix-assisted laserdesorption/ionization time-of-flight mass spectrometry, RapidCommun. Mass Spectrom., 1996, 10(8), 883–888.

14 R. E. Mandrell, L. A. Harden, A. Bates, W. G. Miller, W. F. Haddonand C. K. Fagerquist, Speciation of Campylobacter coli, C. jejuni, C.helveticus, C. lari, C. sputorum, and C. upsaliensis by matrix-assistedlaser desorption ionization-time of flight mass spectrometry, Appl.Environ. Microbiol., 2005, 71(10), 6292–6307.

15 K. L. Wahl, S. C. Wunschel and K. H. Jarman, et al. Analysis ofmicrobial mixtures by matrix-assisted laser desorption/ionization

Analyst, 2011, 136, 1739–1746 | 1745

Publ

ishe

d on

21

Febr

uary

201

1. D

ownl

oade

d by

Uni

vers

ity o

f Pr

ince

Edw

ard

Isla

nd o

n 31

/10/

2014

06:

48:1

5.

View Article Online

time-of-flight mass spectrometry, Anal. Chem., 2002, 74(24), 6191–6199.

16 S. C. Wunschel, K. H. Jarman and C. E. Petersen, et al. Bacterialanalysis by MALDI-TOF mass spectrometry: an inter-laboratorycomparison, J. Am. Soc. Mass Spectrom., 2005, 16(4), 456–462.

17 M. J. Donohue, J. M. Best, A. W. Smallwood, M. Kostich,M. Rodgers and J. A. Shoemaker, Differentiation of Aeromonasisolated from drinking water distribution systems using matrix-assisted laser desorption/ionization-mass spectrometry, Anal. Chem.,2007, 79(5), 1939–1946.

18 P. A. Demirev, A. B. Feldman, P. Kowalski and J. S. Lin, Top-downproteomics for rapid identification of intact microorganisms, Anal.Chem., 2005, 77(22), 7455–7461.

19 C. K. Fagerquist, B. R. Garbus and W. G. Miller, et al. Rapididentification of protein biomarkers of Escherichia coli O157:H7 bymatrix-assisted laser desorption ionization-time-of-flight-time-of-flight mass spectrometry and top-down proteomics, Anal. Chem.,2010, 82(7), 2717–2725.

20 C. K. Fagerquist, B. R. Garbus, K. E. Williams, A. H. Bates, S. Boyleand L. A. Harden, Web-based software for rapid top-down proteomicidentification of protein biomarkers, with implications for bacterialidentification, Appl. Environ. Microbiol., 2009, 75(13), 4341–4353.

21 C. K. Fagerquist (CA, USA), L. A. Harden (Richmond, CA, USA),B. R. Garbus (Santa Clara, CA, USA), Inventor; The United Statesof America, as represented by the Secretary of Agriculture(Washington, DC, USA), assignee, Rapid Identification of Proteinsand their Corresponding Source Organisms by Gas PhaseFragmentation and Identification of Protein Biomarkers, 2010.

22 B. Michel, After 30 years of study, the bacterial SOS response stillsurprises us, PLoS Biol., 2005, 3(7), e255.

23 C. K. Fagerquist, Rapid identification of protein biomarkers andtoxins from food-borne pathogens by top-down proteomics, 239thAmerican Chemical Society National Meeting & Exposition, SanFrancisco, CA, 2010.

24 G. Grotiuz, A. Sirok, P. Gadea, G. Varela and F. Schelotto, Shigatoxin 2-producing Acinetobacter haemolyticus associated witha case of bloody diarrhea, J. Clin. Microbiol., 2006, 44(10), 3838–3841.

1746 | Analyst, 2011, 136, 1739–1746

25 M. Miyahara and H. Konuma, Escherichia coli O157 strains whichcaused Japanese outbreaks have residues of bacteriophagesequences, Biol. Pharm. Bull., 1999, 22(12), 1372–1375.

26 M. Lin, J. M. Campbell, D. R. Mueller and U. Wirth, Intact proteinanalysis by matrix-assisted laser desorption/ionization tandem time-of-flight mass spectrometry, Rapid Commun. Mass Spectrom., 2003,17(16), 1809–1814.

27 S. S. Thakur and P. Balaram, Rapid mass spectral identification ofcontryphans. Detection of characteristic peptide ions byfragmentation of intact disulfide-bonded peptides in crude venom,Rapid Commun. Mass Spectrom., 2007, 21(21), 3420–3426.

28 M. Mormann, J. Eble and C. Schwoppe, et al. Fragmentation ofintra-peptide and inter-peptide disulfide bonds of proteolyticpeptides by nanoESI collision-induced dissociation, Anal. Bioanal.Chem., 2008, 392(5), 831–838.

29 H. Lioe and R. A. O’Hair, A novel salt bridge mechanism highlightsthe need for nonmobile proton conditions to promote disulfide bondcleavage in protonated peptides under low-energy collisionalactivation, J. Am. Soc. Mass Spectrom., 2007, 18(6), 1109–1123.

30 M. D. Jones, S. D. Patterson and H. S. Lu, Determination of disulfidebonds in highly bridged disulfide-linked peptides by matrix-assistedlaser desorption/ionization mass spectrometry with postsourcedecay, Anal. Chem., 1998, 70(1), 136–143.

31 J. J. Gorman, B. L. Ferguson, D. Speelman and J. Mills,Determination of the disulfide bond arrangement of humanrespiratory syncytial virus attachment (G) protein by matrix-assisted laser desorption/ionization time-of-flight massspectrometry, Protein Sci., 1997, 6(6), 1308–1315.

32 C. K. Fagerquist, B. R. Garbus, K. E. Williams, A. H. Bates andL. A. Harden, Covalent attachment and dissociative loss ofsinapinic acid to/from cysteine-containing proteins from bacterialcell lysates analyzed by MALDI-TOF-TOF mass spectrometry, J.Am. Soc. Mass Spectrom., 2010, 21(5), 819–832.

33 C. K. Fagerquist, K. E. Williams, A. H. Bates, Identification offoodborne bacteria by high energy collision-induced dissociation oftheir protein biomarkers by MALDI tandem-time-of-flight massspectrometry, paper presented at: 55th American Society of MassSpectrometry Conference, Indianapolis, Indiana, 2007.

This journal is ª The Royal Society of Chemistry 2011