Báo cáo khoa học: Total chemical synthesis and NMR characterization of the glycopeptide tx5a, a heavily post-translationally modified conotoxin, reveals that the glycan structure is a-D-Gal-(1fi3)-a-D-GalNAc pot

Grey Craig1 1 The Clayton Foundation Laboratories for Peptide Biology and2Laboratory for Molecular and Cell Biology, The Salk Institute, La Jolla, CA, USA;3Department of Chemistry, Swedi

Total chemical synthesis and NMR characterization of the glycopeptide tx5a, a heavily post-translationally modified conotoxin, reveals that the glycan structure is a- D -Gal-(1fi3)-a- D -GalNAc

James Kang1,*, William Low1,*, Thomas Norberg3, Jill Meisenhelder2, Karin Hansson4, Johan Stenflo4, Guo-Ping Zhou5,6, Julita Imperial7, Baldomero M Olivera7, Alan C Rigby5,6and A Grey Craig1

1

The Clayton Foundation Laboratories for Peptide Biology and2Laboratory for Molecular and Cell Biology, The Salk Institute,

La Jolla, CA, USA;3Department of Chemistry, Swedish University of Agricultural Sciences, Uppsala, Sweden;4Department of Clinical Chemistry, University of Lund, Malmo General Hospital, Malmo, Sweden;5Center for Hemostasis and Thrombosis Research, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, USA; 6 Marine Biological Laboratory, Woods Hole, MA, USA; 7 Department of Biology, University of Utah, Salt Lake City, UT, USA

The 13-amino acid glycopeptide tx5a

(Gla-Cys-Cys-Gla-Asp-Gly-Trp*-Cys-Cys-Thr*-Ala-Ala-Hyp-OH, where

Trp*¼ 6-bromotryptophan and Thr* ¼

Gal-GalNAc-threonine), isolated from Conus textile, causes hyperactivity

and spasticity when injected intracerebral ventricularly into

mice It contains nine post-translationally modified residues:

four cysteine residues, two c-carboxyglutamic acid residues,

and one residue each of 6-bromotryptophan,

4-trans-hydroxyproline and glycosylated threonine The chemical

nature of each of these has been determined with the

exception of the glycan linkage pattern on threonine and the

stereochemistry of the 6-bromotryptophan residue Previous

investigations have demonstrated that tx5a contains a

disaccharide composed of N-acetylgalactosamine (GalNAc)

and galactose (Gal), but the interresidue linkage was not

characterized We hypothesized that tx5a contained the

T-antigen, b-D-Gal-(1fi3)-a-D-GalNAc, one of the most

common O-linked glycan structures, identified previously in another Conus glycopeptide, contalukin-G We therefore utilized the peracetylated form of this glycan attached to Fmoc-threonine in an attempted synthesis While the result-ing synthetic peptide (Gla-Cys-Cys-Gla-Asp-Gly-Trp*-Cys-Cys-Thr*-Ala-Ala-Hyp-OH, where Trp* ¼6-bromotrypto-phan and Thr*¼ b-D-Gal-(1fi3)-a-D-GalNAc-threonine) and the native peptide had almost identical mass spectra, a comparison of their RP-HPLC chromatograms suggested that the two forms were not identical Two-dimensional1H homonuclear and13C-1H heteronuclear NMR spectroscopy

of native tx5a isolated from Conus textile was then used to determine that the glycan present on tx5a indeed is not the aforementioned T-antigen, but rather a-D-Gal-(1fi3)-a-D -GalNAc

Keywords: Conus textile; glycopeptide synthesis

The diverse array of peptides isolated from the venom of cone snails are known collectively as conotoxins or cono-peptides (if they lack a disulfide-bonded architecture) The growing interest in these peptides stems from their ability to bind receptors and ion channels with high selectivity and unparalleled specificity A distinct feature of most conotox-ins is their relatively small size (10–35 amino acid residues) combined with the presence of a high proportion of cysteine residues that are involved in disulfide bridging [1] In addition, many of the amino acid residues present in conotoxins have undergone post-translational modification; among the diverse array of modifications characterized

to date are glutamic acidfic-carboxyglutamic acid [2], prolinefi4-trans-hydroxyproline [3], tryptophanfi6-L -bromotryptophan [4] and threonine/serinefiO-linked gly-cosylated threonine/serine [5–7]

Conotoxin tx5a (or e-TxIX), which was purified recently

by two independent laboratories from the venom of the mollusc-hunting cone snail, Conus textile, the
cloth-of-gold cone, is comprised of an unusually large number of amino acids that are post-translationally modified [8–10] Uniquely, this 13-amino acid peptide contains four post-translational modifications and two disulfide bonds In

Correspondence to T Norberg, Department of Chemistry, Swedish

University of Agricultural Sciences, SE-750 07 Uppsala, Sweden.

Fax: + 46 18 673476, Tel.: + 46 18 671578,

E-mail: thomas.norberg@kemi.slu.se or A C Rigby, Center for

Hemostasis and Thrombosis, Beth Israel Deaconess Medical Center,

Harvard Medical School, Boston, MA 02115, USA.

Fax: + 1 617 975 5505, Tel.: + 1 617 667 0637,

E-mail: arigby@bidmc.harvard.edu

Abbreviations: DI, deionized; DIPEA, N,N¢-diisopropylethylamine;

DMF, N,N-dimethylformamide; EDT, 1,2-ethanedithiol; Gal,

galac-tose; GalNAc, N-acetyl galactosamine; NMP, N-methylpyrrolidone;

Gla, c-carboxyglutamic acid, HBTU,

O-(benzotriazol-1-yl)-N,N,N¢,N¢-tetramethyluronium hexafluorophosphate; Hypro,

4-trans-hydroxyproline; MTBE, methyl tert-butyl ether; TBTU,

O-(benzotriazol-1-yl)-N,N,N¢,N¢-tetramethyluronium

tetrafluorobo-rate; TCEP, tris-(2-carboxyethyl)-phosphine hydrochloride;

TPPI, time-proportional phase incrementation.

*Note: These authors contributed equally to the work.

(Received 9 August 2004, revised 20 October 2004,

accepted 27 October 2004)

total, nine of the 13 residues in tx5a are modified, making it

one of the most highly modified gene products identified to

date The native peptide was purified to apparent

homo-geneity, and was reported to cause tremors, hyperactivity

and spastic gait when injected intra-cerebral ventricularly

[8] An underlying mechanism for the biological activity of

the tx5a peptide was proposed by Rigby et al who

suggested that tx5a may target presynaptic Ca2+channels

or that it might act on these channels via other

mecha-nisms, such as through G-protein-coupled presynaptic

receptors [9]

The composition of the O-glycan on the threonine residue

of tx5a was previously investigated [9] It was shown that the

peptide contained N-acetylgalactosamine (GalNAc) and

galactose (Gal) in approximately equal molar amounts;

however, the anomeric stereochemistry and the glycan

linkages were not determined We have previously

charac-terized contulakin-G, a glycopeptide isolated from Conus

geographusvenom, and determined that this glycopeptide

possessed the same monosaccharide constituents as tx5a

We further demonstrated that these monosaccharides were

linked in the b-D-Gal-(1fi3)-a-D-GalNAc configuration of

the T-antigen [6]

Here we report the synthesis of a peptide identical

in composition to that of tx5a, using a racemic D/L

-6-bromotryptophan derivative and a Thr10 derivative

carrying a b-D-Gal-(1fi3)-a-D-GalNAc glycan substituent

However, the peptide synthesized proved to be disparate

from native tx5a isolated from the Conus textile venom To

better understand the incongruence of these peptides we

reinvestigated the glycan linkage configuration of the

isolated and purified native tx5a venom, using both1H-1H

homonuclear and 13C-1H heteronuclear two-dimensional

NMR spectroscopy The data clearly indicated that the tx5a

glycan is in an a-D-Gal-(1fi3)-a-D-GalNAc configuration

Taken together, these data demonstrate that two Conus

glycopeptides identified to date possess the same

monosac-charide constituents within their glycan, Gal and GalNAc,

but their interresidue linkages are different (alpha vs beta)

The results suggest that O-glycosylation in Conus peptides is

likely to be more complex than had originally been expected

and highlights another post-translational modification that

the Conus species employ to adapt to their ever changing environment

Experimental procedures Peptide synthesis

We carried out both manual and automated syntheses as described below, using an Fmoc solid-phase strategy with the amino acid derivatives shown in Scheme 1 The manual synthesis was carried out on an Fmoc-Hypro Wang resin (Chem-Impex, Wood Dale, IL, USA; 0.4 g, 0.7 mmolÆg)1) Each cycle consisted of Fmoc deprotection with 20% piperidine in N-methylpyrrolidone (NMP), followed by Fmoc amino acid coupling using O-(benzotriazol-1-yl)-N,N,N¢,N¢-tetramethyluronium tetrafluoroborate (TBTU) and N,N¢-diisopropylethylamine (DIPEA) in NMP To avoid diketopiperazine formation, Fmoc-Ala-Ala (Bachem, Torrance, CA, USA) was coupled as the first Fmoc amino acid A two-fold excess of Fmoc amino acids was used in the coupling reactions with the exception of Fmoc-D/L -6-bromotryptophan and per-O-acetylated Fmoc [b-D -Gal-(1fi3)-a-D-GalNAc]-Thr [7] where a 20% excess was used The efficiency of the coupling reactions was checked using the Kaiser ninhydrin test The dried peptide resin (0.47 g) was treated with 4.5 mL of trifluoracetic acid in the presence

of 250 lL thioanisole, 125 lL 1,2-ethanedithiol (EDT), and

125 lL deionized (DI) water at room temperature for 1.5 h After precipitation and washing of the cleaved peptide with cold methyl tert-butyl ether (MTBE, 2· 20 mL), the peptide was taken up in 0.1% aqueous trifluoroacetic acid and 60% acetonitrile (2· 10 mL)

Automated chemical synthesis was performed on an ABI 432A peptide synthesizer (Applied Bioysystems, Foster City, CA, USA) employing O-(benzotriazol-1-yl)-N,N, N¢,N¢-tetramethyluronium hexafluorophosphate (HBTU)/ DIPEA/DMF for coupling and piperidine/DMF for Fmoc deprotection Coupling of Fmoc-Ala-Ala to Fmoc-Hypro Wang resin (50 mg) was performed manually and then loaded onto the automated synthesizer for the remainder of the sequence Three-fold excess of amino acid derivatives were used in the coupling reactions with the exception of

per-Scheme 1 Sequence of addition of amino acid derivatives during the solid-phase glycopeptide synthesis The amino acids are numbered starting from the amino terminal according to accepted nomenclature As solid-phase pep-tide synthesis starts from the carboxy terminal, the order of addition is from higher to lower number.

O-acetylated Fmoc [b-D-Gal-(1fi3)-a-D-GalNAc]-Thr

where a 10% excess was used The peptide synthesizer used

conductance monitoring to check the efficiency of the

coupling reactions In order to scale up the automated

synthesis to the same level as the manual synthesis

(280 lmol) we carried out nine separate automated

synthe-ses Each dried peptide-resin (65 mg· nine aliquots ¼ total

weight, 585 mg) was treated with 900 lL of trifluoracetic

acid in the presence of 50 lL thioanisole, 25 lL EDT and

25 lL DI water at room temperature for 1.5 h After

precipitation and washing of the cleaved peptide with cold

MTBE (2· 5 mL), the nine precipitates were collected and

the peptide was taken up in 0.1% aqueous trifluoracetic acid

and 60% acetonitrile (2· 5 mL)

Purification of the manual synthesis

[per-O-acetyl-b-D-Gal-(1fi3)-a-D-GalNAc-Thr10, Cys(t-butyl thiol)2-8,

Cys(Acm)3-9]-tx5a crude product on an analytical HPLC

(1% per min gradient with 0.1% aqueous trifluoroacetic

acid as buffer A and 60% acetonitrile in 0.1% aqueous

trifluoroacetic acid as buffer B) gave two major components

whose observed mass, using ESI-MS, was consistent with

the expected product A similar result was obtained from the

automated synthesis The first component (
hydrophilic)

eluted at
46% acetonitrile, the second (
hydrophobic)

eluted at
48% acetonitrile Because of the low yields from

each synthesis, the material was combined for the following

treatments We estimated that prior to summation, the yield

from automated and manual synthesis were approximately

equivalent The crude extract was loaded onto a

45· 320 mm column packed with Vydac C18 15–20 lm

particles and eluted using a preparative HPLC (PrepLC/

System 6000, Waters Corporation, Millford, MA, USA)

equipped with a gradient controller, a variable wavelength

detector (Waters, model 486) and Waters 1000 PrepPack

cartridge chamber in 0.1% aqueous trifluoracetic acid, using

a gradient of 60% acetonitrile in 0.1% aqueous

trifluorace-tic acid Each component was injected on analytrifluorace-tical HPLC

under isocratic conditions to check for purity and quantity

Approximately 240 nmol of the hydrophilic component

(eight aliquots at 30 nmol) and 150 nmol of hydrophobic

component (five aliquots at 30 nmol) were lyophilized for

the sugar de-O-acetylation reaction Each dried aliquot was

treated with 500 lL 150 mM NaOCH3 in methanol for

20 min at 25C and then quenched with 200 lL DI water

Purification of the [b-D-Gal-(1fi3)-a-D-GalNAc-Thr10,

Cys(t-butyl thiol)2,8, Cys(Acm)3,9] hydrophilic and

hydro-phobic products on preparative HPLC identified

hydrophi-lic and hydrophobic components with an observed mass of

2252.2 m/z (ESI-MS), which are consistent with the

theor-etical peptide mass (2252.6 Da)

Disulfide bond formation

In preparation for the Cys2-8 disulfide reaction, each

component was injected on analytical HPLC under

isocratic conditions to check for purity and quantity

Approximately 160 nmol of the hydrophilic component

(20 aliquots at 8 nmol) and 120 nmol of hydrophobic

component (20 aliquots at 6 nmol) were lyophilized Each

dried aliquot was treated with 0.17M citric acid (750 lL,

pH 6.5) and 1M tris-(2-carboxyethyl)-phosphine

hydro-chloride (TCEP) (150 lL) at 37C for 180 min, and then

quenched with 0.1% aqueous trifluoracetic acid (500 lL)

In order to minimize complications that result from these Gla-containing peptides forming divalent metal ion com-plexes (i.e creating peak broadening or multiple peaks when analyzed on HPLC), 1% CaCl2 (100 lL) was added to each aliquot Purification of the [b-D -Gal-(1fi3)-a-D-GalNAc-Thr10, Cys2-8, Cys(Acm)3-9] hydro-philic and hydrophobic products by analytical HPLC indicated hydrophilic and hydrophobic components whose observed mass (ESI-MS) were consistent with the calculated peptide mass To test for completion of the Cys2-8 disulfide bridge formation, 20 mM K3Fe(CN)6 (15 lL) was added to 30 lL (1 nmol) of the hydrophilic component (pH 7) at 25C for 20 min, and then the pH was readjusted to 5 using 50% aqueous acetic acid Coinjection of the untreated and K3Fe(CN)6-treated hydrophilic components on analytical HPLC indicated a difference in retention time indicative of formation of the disulfide bridge which was confirmed with ESI-MS analysis

Following the injection of both hydrophilic and hydro-phobic components onto an analytical HPLC column under isocratic conditions to ensure the purity and quantity of each peptide,
120 nmol of the hydrophilic component (30 aliquots at 4 nmol) and 80 nmol of hydrophobic com-ponent (20 aliquots at 4 nmol) were lyophilized for a

Cys3-9 disulfide reaction Each dried aliquot was dissolved with 0.1% aqueous trifluoroacetic acid (400 lL) and 40 lL 1% CaCl2at 0C, and then treated with 2 lL of 0.1% iodine

in methanol at 25C for 15 min Finally, 2 lL of 2.5% ascorbic acid in DI water was added to quench the reaction and eliminate excess iodine

HPLC purification of cyclo tx5a Purification of the cyclo 2-8, 3-9[b-D-Gal-(1fi3)-a-D -GalNAc-Thr10]-tx5a hydrophilic and hydrophobic prod-ucts on analytical HPLC (10 mm· 250 mm Vydak C18

300 A˚ pore size) with 0.1% aqueous trifluoroacetic acid as buffer A and 60% acetonitrile in 0.1% aqueous trifluoro-acetic acid as buffer B (gradient 1% per min) resulted in components whose observed masses (ESI-MS) were consis-tent with the expected peptide masses Each component was injected on an analytical HPLC under gradient conditions

to check for purity and quantity The hydrophilic compo-nent was collected at
20% acetonitrile (174 lg, 90 nmol) The hydrophobic component was collected at approxi-mately 23% acetonitrile (66 lg, 34 nmol) ESI and matrix assisted laser desorption mass spectrometry (MALDI-MS) measurement of both components resulted in intense species consistent with the correct product (see below) Both the hydrophilic and hydrophobic components were found to be 99% pure as assessed using an orthogonal ion pairing agent system (triethylammonium phosphate, pH 2.3 as buffer A, 60% acetonitrile as buffer B with a gradient from 10 to 50%

B in 40 min)

Enzyme hydrolysis Approximately 1 nmol of native tx5a, tx5a hydrophilic and tx5a hydrophobic were incubated with 25 mU b-galactosi-dase from bovine testes (Glyko, Inc., Novato, CA, USA) in

100 lL of 100 mMsodium citrate/phosphate pH 4 at 32C

for 24 h As a positive control of enzyme activity,

contul-akin-G, a b-D-Gal-(1fi3)-a-D-GalNAc containing

glyco-peptide, and native tx5a were simultaneously incubated and

reacted with the same vial of the enzyme b-galactosidase

Native tx5a, tx5a hydrophilic, and tx5a hydrophobic (each

1 nmol) were incubated with 2.5 mU of endo-O-glycosidase

(endo-a-N-acetylgalactosaminidase) (Prozyme, Inc., San

Leandro, CA, USA) in 50 lL of 50 mMNaHPO4pH 5 at

32C for 24 h As a positive control of enzyme activity,

contulakin-G and native tx5a were coincubated with the

enzyme endo-O-glycosidase In each case, the enzyme

reactions were stopped with addition of 10 lL 10%

aqueous trifluoroacetic acid and immediately injected onto

RP-HPLC and fractions collected were collected and

analyzed with ESI and MALDI-MS

Chemical reduction

The native and synthetic tx5a (hydrophilic and

hydropho-bic) were incubated with 50 mMTCEP for 30 min at 32C

prior to injection on reverse-phase HPLC, collection and

analysis with ESI-MS

Coelution

The native and synthetic tx5a (hydrophilic and

hydropho-bic) were coinjected onto a Vydac C18RP-HPLC column

(2.1· 150 mm) and eluted with a 1% per min gradient

from 0% B to 45% B (where buffer A was 0.055% aqueous

trifluoroacetic acid and buffer B was 0.05% trifluoroacetic

acid in 90% aqueous acetonitrile)

Mass spectrometry

HPLC purified fractions were analyzed with both ESI-MS

and MALDI-MS Samples for electrospray analysis were

diluted 1 : 1 with 1% acetic acid in methanol and infused at

1 lLÆmin)1into an Esquire LC electrospray quadrupole ion

trap mass spectrometer (Bruker Daltonics, Billerica, MA,

USA) Previously, we have demonstrated the mass accuracy

for our electrospray instrument for nonresolved isotopic

clusters of metal chelate complexes to be ± 1.0 m/z when

compared with the calculated average mass Samples for

MALDI-MS analysis were mixed with

a-cyano-4-hydroxy-cinnamic acid and irradiated with 282 nm irradiation from

a nitrogen laser using a DE-Star (Perceptive, Framingham,

MA, USA) mass spectrometer The mass accuracy of the

MALDI instrument for resolved isotopic clusters is ±

0.2 m/z when compared with the calculated monoisotopic

mass

Purification of native tx5a

Native tx5a (e-TxIX) was purified from Conus textile venom

as described previously [9] Briefly, the venom from Conus

textilecone snails was expressed manually The lyophilized

venom extract (200 mg) was dissolved in 0.2Mammonium

acetate and chromatographed on a Sephadex G50 superfine

column (2.5· 92 cm) equilibrated with 0.2M ammonium

acetate buffer, pH 7.5, and eluted with a flow rate of

9.2 mLÆh)1 The column fractions were monitored using

absorption (A) at 280 and 214 nm Column fractions were subjected to direct Gla analysis following alkaline hydrolysis [11,12] The material in the major Gla-containing peak was further purified on a reverse-phase column (HyCrom C18,

5 l; 10· 250 mm) in 0.1% trifluoroacetic acid and eluted with a linear acetonitrile gradient 20–40% B (Buffer A: 0.1% trifluoroacetic acid, water; Buffer B: 0.1% trifluoro-acetic acid, acetonitrile)

NMR spectroscopy Native tx5a NMR samples were dissolved initially in 99.8% D2O and heated to 50C at a neutral pH of 7.0 in the presence of Chelex 100 to ensure that all trace metal ions were removed This sample was then lyophilized and redissolved in 350 lL of 99.96% D2O (0.7 mM) (Cambridge Isotope Laboratories, Andover, MA, USA),

to a noncorrected pH of 5.60 and transferred to a 4 mm NMR tube All spectra were acquired on a Varian Unity INOVA spectrometer with a proton frequency of 499.695 MHz (Varian Inc., Palo Alto, CA, USA) The carrier frequency was set on the water resonance, which was suppressed using presaturation or a wet pulse sequence Preliminary one-dimensional spectra were acquired over a range of temperatures (5–35C) with

16 000 real data points, 256 summed scans and a spectral width of 8000 Hz The final two-dimensional 1H homo-nuclear and 13C-1H heteronuclear correlation data sets were collected at 12C

Two-dimensional NOESY spectra were recorded with mixing times of 150 and 320 ms A total of 2048 (or 4096) real data points were acquired in t2, 512 time-proportional phase increments (or States-TPPI) in t1, with a spectral width of 8000 Hz in the observed (F2) dimension A total of

128 summed scans were collected with a relaxation delay of 1.3 s between scans Spectra were processed with a sine bell window function shifted by 30 in t2 (applied over 1024 points) and a sine bell window function shifted by 30 in t1 (applied over all 512 acquired points) using the Varian processing software, VNMR (Varian Inc., Palo Alto, CA, USA) All data were zero-filled to a 4096 by 2048 matrix using theVNMRprocessing program TOCSY spectra were recorded and processed as described for the NOESY with the exception that 4096 real data points were acquired in t2, with 384 time-proportional phase incrementation (TPPI or States-TPPI) increments in t1 A 35 ms mixing time was used in collecting 256 summed scans employing the

MLEV-17 spinlock sequence A DQF-COSY spectrum was recor-ded with 4096 real t2 points, 64 summed scans, and 712 TPPI increments to ensure increased resolution The spectrum was multiplied by sine bell window functions shifted by 30 in t2 and 30 in t1 and zero-filled to a 2048 by

1024 (real) matrix A two-dimensional13C-1H heteronuclear single quantum coherence (HSQC) spectrum was recorded with 2048 real data points in t2, with 192 time-proportional phase instrumentation (TPPI or States-TPPI) increments in t1 and spectral widths of 8000 Hz and 17 591 Hz in the1H and13C dimensions, respectively A total of 256 summed scans were collected with a relaxation delay of 1.3 s All1H and1H-13C correlation assignments were performed using FELIX2000, which is part of theINSIGHTsuite of programs (Accelrys, San Diego, CA, USA)

After synthesis and deprotection of tx5a from the resin, we

obtained two components with the expected mass, herein

referred to as hydrophilic and hydrophobic Because a

racemic mixture ofL/D6-bromotryptophan was used in the

synthesis to insure that we would synthesize a tx5a analog

corresponding to the native peptide (irrespective of which

6-bromotryptophan isomer was incorporated in the native

tx5a peptide) we propose that the hydrophilic and

hydro-phobic fractions correspond to theL/D6-bromotryptophan

isomers of tx5a In order to further compare these

hydrophilic and hydrophobic fractions the mass spectra of

the synthetic products were determined following reduction

of the disulfide bonds (to remove potential complexity of

data due to the disulfide arrangement) When measured in

the negative ionization mode, the ESI mass spectra of all three samples were almost identical in appearance Figure 1 shows (A) the hydrophilic component and (B) the reduced native tx5a (a similar result was observed for the hydro-phobic component, data not shown) In Fig 1B, three major species observed at m/z 994.2, 972.1 and 950.5 (identified as MR¢, MR¢¢ and MR¢¢¢) were interpreted as corresponding to [MR+Fe-5H]2–, [MR+Fe-CO2-5H]2–and [MR+Fe-2CO2-5H]2–where MRcorresponds to the expec-ted average mass of chemically reduced native tx5a (m¼ 1935.81 Da) As previously proposed [13], the fragment ions are formed in the mass spectrometer from the facile loss of

CO2 (e.g from either of the two c-carboxyglutamic acid residues or other acidic groups) rather than from synthetic by-products based on the RP-HPLC, ion exchange chro-matography and capillary zone electrophoresis results (data

1000

800

600

400

200

0

MR'

MR''

MR'''

m/z

1855 1845

1835 100

Mass (m/z)

800

600

400

200

0

m/z

1855 1845

1835 100

Mass (m/z)

A

B

Fig 1 Electrospray mass spectrum of (A)

chemically reduced synthetic hydrophilic cyclo

2-8, 3-9[6-L/D-bromo-Trp7,

b-D-Gal-(1fi3)-a-D -GalNAc-Thr10]-tx5a compared with (B)

chemically reduced native tx5a where M R ¢ ¼

[M R +Fe-5H] 2– species Insets show the

cor-responding MALDI resolved isotope

distri-butions of the [M -CO -H] – species.

not shown) Other species present in Fig 1 correspond to

sodium cationization (i.e +Na-H) of the intact and

fragment ions Insets in Fig 1 are the MALDI-MS resolved

isotope distribution of the chemically reduced

hydrophi-lic and native samples (the species corresponds to [MR

-2CO2-H]–, observed monoisotopic m/z 1844.5 and 1844.7,

respectively, compared with the calculated monoisotopic

[MR-2CO2-H]–mass of 1844.46 Da)

After selective folding of the disulfide bridges of the

hydrophilic component of tx5a, we observed similar ESI

negative mass spectra from the synthetic hydrophilic and

native tx5a, as shown in Fig 2 (a similar result was observed

for the hydrophobic component, data not shown) In

Fig 2B, the M¢, M¢¢ and M¢¢¢ species observed at m/z 992.1,

970.1 and 948.0 were interpreted as corresponding with

[M+Fe-5H]2–, [M+Fe-CO2-5H]2–and [M+Fe-2CO2-5H]2–

where M corresponds to the expected average mass of native tx5a (m¼ 1931.76 Da) The insets in Fig 2 show the MALDI-MS (negative ion mode), resolved isotope distri-bution measurements for the hydrophilic and native species (observed monoisotopic at m/z 1840.13 and 1840.32, respectively, compared with the calculated monoisotopic [M-2CO2-H]– mass of 1840.42 Da) The mass shift of

4 Da (MR) M) confirms the formation of the two disulfide bridges

However, comparison of the retention times of the hydrophilic tx5a, hydrophobic tx5a and native tx5a (Table 1) reveals that the three peptides have different chromatographic properties and can be clearly distinguished when analyzed under either nonreducing or reducing conditions In particular, RP-HPLC chromatography of chemically reduced native tx5a and the reduced hydrophilic

1500

1000

500

0

m/z

1075

M' M''

M'''

1855 1845

1835 100

Mass (m/z)

1200

600

0

M'

M''

M'''

m/z

1855 1845

1835 100

Mass (m/z)

A

B

Fig 2 Electrospray mass spectrum of (A) synthetic hydrophilic [6-L/D-bromo-Trp7, b- D -Gal-(1fi3)-a- D -GalNAc-Thr10]-tx5a compared with (B) native tx5a where M¢ ¼ [M+Fe-5H]2)species Insets show the cor-responding MALDI resolved isotope distri-butions of the [M-CO -H] – species.

tx5a that were coinjected (Fig 3) reveals a small but

significant difference in the chromatographic retention time

of these two glycopeptides

Similarly, when the chemically synthesized hydrophilic

tx5a or hydrophobic tx5a was incubated with

b-galactosi-dase, the retention time of the product (Table 1) and the

observed mass were altered as a result of the elimination of

the galactose residue as determined by MALDI-MS (data

not shown) In contrast, the retention time of native tx5a

did not change when incubated under these conditions In

order to exclude the possibility that the absence of enzyme

hydrolysis was due to a contaminating enzyme inhibitor

present in the native tx5a preparation, we added a control

glycopeptide (contulakin-G) to this incubation mixture

We observed that the enzyme was able to hydrolyze the

galactose residue on the control glycopeptide (data not

shown) In addition, both the hydrophilic tx5a and

hydrophobic tx5a peptides demonstrated a shifted HPLC

retention time following incubation with

endo-O-glycosi-dase (Table 1) and an observed mass change as a result of

the elimination of the entire glycan In contrast, the

retention time of native tx5a did not change when

incubated under these conditions The presence of

contul-akin-G (positive control) was used to validate the activity

of the endo-O-glycosidase enzyme, which was unable to

hydrolyze native tx5a Together these data suggest that the

glycan configuration of native tx5a is distinct from the

synthetic tx5a peptides and contulakin-G Therefeore,

two-dimensional DQF-COSY, TOCSY, NOESY and HSQC

spectra of the native tx5a glycopeptide were collected in

99.96% D2O at 12C, pD 5.6 These data, in combination

with data collected previously in 90 : 10 H2O/D2O enabled assignment of the amino acid and sugar spin systems of the glycopeptide [9] Interestingly, O-glycosylation of Thr10 perturbed the b-carbon 13C chemical shift (81.8 p.p.m.), which is downfield from the expected chemical shift (67.9– 68.3 p.p.m.) and in support of the glycan linkage at this site [14] Several resonances that were attributed to the glycan moiety of tx5a, localized within a spectral envelope between 3.4 p.p.m and 4.0 p.p.m., remained unassigned following our initial assignment of the glycopeptide backbone and side chain resonances [9] The resonances

of the monosaccharides residues GalNAc and Gal were primarily assigned from DQF-COSY and TOCSY spectra commencing with the anomeric protons at 4.79 and 4.82 p.p.m., respectively (Table 2) Both of these proton resonances demonstrated strong correlation cross-peaks to two additional high field proton signals that were tenta-tively assigned H2 and H3 for the respective monosaccha-rides (Fig 4A) These assignments were confirmed using the single interproton scalar connectivity measured by the DQF-COSY spectrum The remaining GalNAc and Gal proton resonances were assigned using the aforementioned spectra in combination with NOESY data and a natural abundance 13C-1H HSQC spectrum that enabled each carbon to be correlated with its directly bonded proton or protons (Tables 2 and 3)

Strong NOEs between the anomeric and H2 protons and

3J1,2coupling constants of 4.25 Hz for both the GalNAc and Gal monosaccharides identify an a configuration for both anomeric centers within the glycan (Table 2) The3J2,3 coupling constants were 7.92 and 7.84 Hz, respectively, for the GalNAc and Gal monosaccharides Furthermore, the H3 resonance of GalNAc showed a strong NOE to the

Table 1 Comparison of the reverse-phase HPLC retention times (in minutes) of native tx5a, synthetic hydrophilic, synthetic hydrophobic peptides under nonreducing and reducing conditions, and after incubation withb-galactosidase and O-glycosidase.

Retention Time (min)

Chemically reduced Native tx5a Chemically reduced hydrophilic tx5a

Fig 3 Reverse-phase HPLC chromatography of a coinjection of

chemically reduced native tx5a and synthetic reduced hydrophilic cyclo

2-8, 3-9[6- / -bromo-Trp7, b- -Gal-(1fi3)-a- -GalNAc-Thr10]-tx5a.

Table 2 1 H, 13 C chemical shifts and scalar coupling constants for the glycan monosaccharides in tx5a in 99.96% D 2 O at 285.5 K (relative to sodium 2,2-dimethyl-2-silapentane-5-sulfonate).3J x,y , 3-bond coupling constant.

Proton ( 1 H)

1

3

3

3

anomeric proton of Gal, which identified that the Gal

residue is linked to the GalNAc monosaccharide through a

H1–H3 linkage (Fig 4B) Furthermore, the low field C3

(Table 3) carbon chemical shift (77.4 p.p.m.) of the GalNAc

residue supports it being glycosylated at position 3 Taken

together, these data identified the glycan as a-D

-Gal-(1fi3)-a-D-GalNAc

There are several NOEs between the glycan and the

glycopeptide side-chain atoms of tx5a, which suggests that

the monosaccharides are conformationally less flexible,

well ordered and within 5 A˚ of these glycopeptide

side-chains at 12C (Table 4) Specifically, the side-chain

protons of Thr10, Ala12 and Hyp13 interact with the

glycan protons Several NOEs are observed between the anomeric proton of GalNAc and the side-chain atoms of Thr10; Thr10b (strong NOE) and Thr10CH3 (medium NOE) (Fig 5) For Ala12 the Ala12a and Ala12bCH3 side-chain protons demonstrate medium and strong NOEs, respectively, with the N-acetyl CH3 of GalNAc

at 1.79 p.p.m., which may alter the magnetic and chem-ical environment of this moiety and help us to understand this fairly unique chemical shift frequency (Fig 5) In addition, there are several weak NOEs between the N-acetyl CH3and GalNAc H3, GalNAc H4and Gal H3, which further supported a well ordered carbohydrate moiety at 12C (Table 3)

Fig 4 Two-dimensional1H spectra of 0.7 m M tx5a (e-TxIX) collected in 100% D 2 O or 90% : 10% H 2 O/D 2 O, respectively, at 500 MHz (A) TOCSY spectrum collected in 100% D 2 O illustrating the alpha region of the data, which includes the monosaccharide resonances and (B) NOESY spectrum collected in 90% : 10% H 2 O/D 2 O (H 2 O resonance at 4.65 p.p.m.) of this same region collected with a mixing time of 320 ms All data were collected at 12 C Specific carbohydrate resonances are assigned in addition to protons of amino acids residues from the tx5a peptide including Gly6, Thr10 and Pro13 (A) Illustrates the intraresidue carbohydrate assignments GalNAc (GN) and Gal (G), respectively In B, many of these same intraresidue assignments are labeled in addition to the interglycosidic linkage between GNH 3 of GalNAc and GH 1 of Gal, which is labeled in bold The amino acids Gly6, Thr10 and Pro13 are represented by 6G, 10T and 13P.

Table 4 NOEs between the tx5a peptide resonances and the protons (1H) within the Gal-GalNAc disaccharide The nomenclature represents that used in Figs 4 and 5.

Disaccharide: Gal-GalNAc tx5a

Proton ( 1 H)

Chemical shift (p.p.m.) Proton ( 1 H)

Chemical shift (p.p.m.)

Table 3 tx5a Gal-GalNAc NOE interactions and their corresponding

proton (1H) chemical shifts in 99.96% D 2 O at 285.5 K (relative to

sodium 2,2-dimethyl-2-silapentane-5-sulfonate).

Proton ( 1 H)

Chemical

shift (p.p.m.) Proton ( 1 H)

Chemical shift (p.p.m.)

The tx5a peptide from Conus textile has the greatest

diversity of post-translational modifications found in any

conotoxin hitherto characterized There are two disulfide

crosslinks, a hydroxylated proline residue, a brominated

tryptophan residue, and two c-carboxylated glutamic acid

residues In addition, there is an O-glycosylated threonine

residue, where the glycan moiety contains equimolar

amounts of GalNAc and Gal We synthesized the tx5a

peptide with the disulfide connectivity characteristic of

the previously characterized T-superfamily (Cys2-Cys8,

Cys3-Cys9) and assumed that the glycan moiety would

be the T-antigen, as was previously shown for

contula-kin-G, i.e b-D-Gal-(1fi3)-a-D-GalNAc O-linked to

threo-nine [6]

The synthetic strategy is briefly outlined in Scheme 1

During the chemical synthesis we used a selective cysteine

deprotection strategy to obtain the correct disulfide-bonding

pattern In addition, we investigated the relative merits of

manual vs automated Fmoc synthesis of this extremely

complicated target molecule The very low yield obtained,

0.027% (based on our rough estimated 50 : 50 split between

automated and manual synthesis, the 90 nmol of

hydro-philic and 34 nmol of hydrophobic tx5a analogs) is in

contrast with yields (30%) previously obtained for

nondi-sulfide bridge-containing glycopeptides using either manual

or automated strategies [6] We note, however, that even in

the synthesis of nondisulfide bridge-containing

glycopep-tides the yield is dramatically affected by the scale of the

reaction, the excess of amino acids used, and the level of

purity desired Here, our reaction scale was limited by the costs of the reagents and our desire to obtain peptides that were of the highest purity Also, the use of only a slight excess (10–20%) of some expensive amino acids contributed

to the low yield of the desired product, and increased the formation of truncated products In summary, by using a selective Cys deprotection strategy we successfully obtained the desired disulfide connectivity, but this may have partially contributed to the very low yields We note also that determination of the stereochemistry of the 6-bromotryp-tophan residue as either L or D, and utilization of the appropriate resolved precursor would result in a signifi-cantly improved yield

Surprisingly, the chemically synthesized peptides did not coelute with the native peptide as demonstrated by RP-HPLC The difference between native and synthetic peptides is most probably associated with the configuration

of the glycan moiety attached to Thr10 We demonstrated that the glycan of the synthetic peptide could be hydrolyzed

by b-galactosidase, as well as by endo-O-glycosidase, as one would expect for the glycan in a T-antigen configuration These enzymes were previously shown to also hydrolyze the glycan moiety of contulakin-G [6] However, the native tx5a peptide was not amenable to hydrolysis by these two glycosidases The failure to hydrolyze the native peptide was not due to the presence of an inhibitor in the native preparation as demonstrated when native and synthetic peptides were mixed The synthetic peptide was cleaved by the glycosidases, while the native peptide was resistant These data support that the intrinsic carbohydrate proper-ties of the glycan moieproper-ties linked to these peptides are distinct and more importantly that the tx5a glycan is comprised of interglycosidic linkages that are not recognized and thus not cleaved by these enzymes These data permit us

to conclude that, contrary to our expectations and prior results with contulakin-G [6], the glycan present on the tx5a peptide is not the T-antigen

Initial investigations by our laboratory identified that GalNAc and Gal are present in equivalent concentrations, but we did not further determine the configuration of this carbohydrate However, the different elution profiles of the native tx5a and the synthetic peptides constructed with the carbohydrate in the T-antigen configuration combined with the inability of the aforementioned glycosidic enzymes to hydrolyze the tx5a glycan linked to Thr10 identifies that the difference must be attributable to the interglycosidic linkage

of the native tx5a glycan, which is clearly not in a typical T-antigen configuration To better characterize the confi-guration of this glycan moiety we used standard two-dimen-sional homonuclear and heteronuclear (natural abundance) NMR spectroscopy Using the information gleaned from our two-dimensional COSY, NOESY and 13C-HSQC experiments we assigned the1H and13C chemical shifts of all nuclei with the exception of those that remained spectrally degenerate The anomeric protons identified in the COSY and TOCSY spectra (collected at 12C) at 4.79 p.p.m and 4.82 p.p.m (GalNAc and Gal, respectively) provided a good starting place for the through-bond scalar assignment within each sugar moiety (Fig 4A) Most spectral degeneracy was resolved through the use of a13 C-HSQC experiment and the assignments completed using NOESY spectra collected at several mixing times (Fig 4B)

Fig 5 A region of the 500 MHz 2D NOESY (320 ms) collected in

90% : 10% H 2 O/D 2 O illustrating the NOEs observed between the

amino acids side chains of residues Thr10 and Ala12 and the

carbohy-drate moieties of GalNAc (GN) and Gal (G) The individual amino

acids Thr10, Ala12 are represented by 10T and 12A, respectively, while

GNH 1 and GNH 3 represents the H1 and H2 protons from GalNAc,

and GNCH 3 represents the methyl group (CH 3 ) that is within the

GalNAc acetyl group.

These13C data also identified that the b-carbon of Thr10,

was shifted to lower field (81.8 p.p.m.), which supported

that Thr10 was the glycosylation site (as we already

believed) The anomeric configuration and interglycosidic

linkage patterns were identified using several through-bond

scalar measurements, the3J1,2coupling constant between

the anomeric (H1) and H2 protons (1H’s) of each

carbo-hydrate moiety and resonance assignments Specifically, the

small 3J1,2 and larger 3J2,3 scalar coupling constants

identified that both carbohydrate moieties were in an alpha

configuration In addition, the low field chemical shift of the

C3-carbon of the GalNAc (77.4) strongly supported this

carbon as the interglycosidic linkage carbon Together these

data suggested that the interglycosidic linkage between

GalNAc and Gal was 1–3 in the alpha configuration These

data were confirmed by the strong NOE between the1H at

position 3 (H3) of GalNAc and the anomeric (H1)1H of

Gal This linkage pattern helps us to better understand the

resistance to hydrolysis by the aforementioned glycosidic

enzymes, while identifying a linkage pattern that is disparate

from that previously identified for contulakin-G [6]

Inter-estingly, several additional NOEs were identified between

the glycan linked to Thr10 and other tx5a residues as

illustrated in Fig 5 These NOEs support that the

carbo-hydrate moieties interact with the glycopeptide, suggesting

that the carbohydrate is conformationally well structured

This apparent reduction in conformational flexibility (on

the NMR time scale) has been identified previously in other

glycosylated peptides and may further support a functional

role of the glycan in receptor-mediated function although

this requires further investigation It was completely

unex-pected that the only two characterized Conus peptides

containing the same sugar moieties attached to the same

aglycone residue, Thr, would have different configurations

This strongly suggests that the post-translational enzymes

necessary to catalyze O-glycosylation of threonine residues

are different for Conus geographus (contulakin-G) and

Conus textile (tx5a) venoms This conclusion raises a

number of additional questions that necessitate further

investigation

Specifically, what is the actual structure of the glycan

moiety in the native tx5a peptide? Our NMR data indicates

an a-D-Gal-(1fi3)-a-D-GalNAc-Thr structure for this

gly-can, and a renewed total synthesis effort is currently under

way to confirm this finding Apart from the question

pertaining to the glycan configuration there are more

general and intriguing questions related to the

O-glycosy-lation differences of these peptides Recent studies have

demonstrated that for some Conus peptide

post-transla-tional modifications (such as for the conantokin peptide

family which are all c-carboxylated), a recognition signal

sequence present in the precursor sequence serves as a

binding site to recruit the appropriate enzyme that is

necessary for a specific post-translational modification [15]

One possibility is that different recognition signals in the

tx5a and contulakin-G precursors recruit different glycosyl

transferases

An alternative explanation is centered on the fact that

these peptides belong to different peptide superfamilies that

may process peptides through specific and distinct secretory

pathways Thus, enzymes that carry out the glycosylation to

give the configuration of the T-antigen might be packaged

in the secretory pathway of the contulakin family, but a different set of enzymes may be packaged into the secretory pathway for the T-superfamily of peptides to which tx5a belongs It is also feasible that these two Conus species have taken advantage of this post-translational modification in unique ways that allows them to accommodate evolutionary and environmental changes that are specific for each species These data demonstrate the feasibility of chemically synthesizing peptides, such as tx5a, that possess multiple post-translational modifications This synthesis in and of itself is a significant achievement in lieu of the complexity and number of post-translationally modified amino acids included However, our synthetic efforts and subsequent enzymatic degradation and NMR spectroscopy studies, have revealed that the glycan configuration is not the same

as that previously discovered and reported for contulakin-G (Conus geographus) This surprising result establishes that the O-glycosylation of serine and threonine residues in Conuspeptides are likely to be more complex than had originally been anticipated, involving more than one specialized post-translational modification enzyme Acknowledgements

This work was supported by the National Institutes of Health (GM48677) (B.M.O.), the National Science Foundation (A.C.R.) and conducted in part by the Foundation for Medical Research (A.G.C.).

We would like to thank Jean E Rivier and Josef Gulyas for stimulating conversations and helpful advice.

References

1 Craig, A.G., Bandyopadhyay, P & Olivera, B.M (1999) Post-translationally modified neuropeptides from Conus venoms Eur.

J Biochem 264, 271–275.

2 McIntosh, J.M., Olivera, B.M., Cruz, L.J & Gray, W.R (1984) Gamma-carboxyglutamate in a neuroactive toxin J Biol Chem.

259, 14343–14346.

3 Stone, B.L & Gray, W.R (1982) Occurrence of hydroxyproline in

a toxin from the marine snail Conus geographus Arch Biochem Biophys 216, 756–767.

4 Craig, A.G., Jimenez, E.C., Dykert, J., Nielsen, D.B., Gulyas, J., Abogadie, F.C., Porter, J., Rivier, J.E., Cruz, L.J., Olivera, B.M.

& McIntosh, J.M (1997) A novel post-translational modification involving bromination of tryptophan Identification of the residue, L-6-bromotryptophan, in peptides from Conus imperialis and Conus radiatus venom J Biol Chem 272, 4689–4698.

5 Craig, A.G., Zafaralla, G., Cruz, L.J., Santos, A.D., Hillyard, D.R., Dykert, J., Rivier, J.E., Gray, W.R., Imperial, J., DelaCruz, R.G., Sporning, A., Terlau, H., West, P.J., Yoshikami, D & Olivera, B.M (1998) An O-glycosylated neuroexcitatory Conus peptide Biochemistry 37, 16019–16025.

6 Craig, A.G., Norberg, T., Griffin, D., Hoeger, C., Akhtar, M., Schmidt, K., Low, W., Dykert, J., Richelson, E., Navarro, V., Macella, J., Watkins, M., Hillyard, D., Imperial, J., Cruz, L.J.

& Olivera, B.M (1999) Contulakin-G, an O-glycosylated invertebrate neurotensin J Biol Chem 274, 13752–13759.

7 Lu¨ning, B., Norberg, T & Tejbrant, J (1989) Synthesis of mono-and disaccharide amino acid derivatives for use in solid phase peptide synthesis Glycoconj J 6, 5–19.

8 Walker, C., Steel, D., Jacobsen, R.B., Lirazan, M.B., Cruz, L.J., Hooper, D., Shetty, R., DelaCruz, R.C., Nielsen, J.S., Zhou, L., Bandyopadhyay, P., Craig, A & Olivera, B.M (1999) The T-superfamily of conotoxins J Biol Chem 274, 30664–30671.