Key role of the loop connecting the two beta strands of mussel
defensin in its antimicrobial activity
Bernard Romestand
1
, Franck Molina
2
,Ve
´
ronique Richard
1
, Philippe Roch
1
and Claude Granier
2
1
DRIM, Universite
´
Montpellier 2, France;
2
Centre de Biotechnologie et Pharmacologie pour la Sante
´
, CNRS UMR Montpellier,
France
To elucidate the structural features of the mussel defensin
MGD1 required for antimicrobial activity, we synthesized a
series of peptides corresponding to the main known secon-
dary structures of the molecule and evaluated their activity
towards Gram-positive and Gram-negative bacteria, and
filamentous fungi. We found that the nonapeptide corres-
ponding to residues 25–33 of MGD1 (CGGWHRLRC)
exhibited bacteriostatic activity once it was cyclized by a

been reported [2,3]. Defensins are antimicrobial peptides
isolated from mammals [4], arthropods [5,6], plants [7,8]
and more recently from molluscs [9,10]. They are cationic
molecules belonging to the cysteine-rich family of anti-
microbial peptides. Mammalian defensins comprise human
neutrophil peptides (HPN-1–4), human defensins (HD-5
and 6), two human b defensins (HBD-1 and 2) [11–13] and a
cyclic rhesus theta defensin (RTD-1) [14]. Although all
defensins display antibacterial activity, mammalian and
other vertebrate defensins are quite different from the
arthropod/mollusc defensins in terms of both sequence and
structure [15–17].
MGD1 is a defensin of 39 residues, which has been
isolated from plasma and haemocytes of the edible Medi-
terranean mussel, Mytilus galloprovincialis [10,18]. MGD1
shares the so-called cysteine-stabilized alpha-beta motif
(Csab) with arthropod defensins [19], but it is characterized
by the presence of an additional disulfide bond. The three-
dimensional solution structure of MGD1 has been estab-
lished using
1
H-NMR and mainly consists of a helical part
(residues 7–16) and two antiparallel b-strands (residues
20–25 and 33–39) [16]. The a-helix and the b1-strand are
connected by a distorted type II turn (loop 2), whereas the
loop connecting both strands of the b-sheet (residues 25–33)
includes a type III¢ turn (loop 3) and points out of the core
of the protein.
There is a consensus view that defensins act by disrupting
the cytoplasm membrane [20–24], although the exact mode of

cal C18 HPLC column and peptide integrity was checked by
MALDI-TOF mass spectrometry. The N-terminal residue
of every peptide was blocked by acetylation, and the
C-terminal residue was amidated. Disulfide bond formation
in cysteine-containing peptides was performed by dilution of
the peptide in 20% dimethylsulfoxide, 0.05
M
ammonium
acetate, pH 7.5, for 24 h at room temperature under
agitation [27]. Peptide Q, which comprises two disulfide
bonds, was obtained by sequential formation of the Cys25-
Cys33 disulfide bond, as described above, then removal of
the acetamidomethyl group (0.1
M
iodine in water acidified
to pH 4 with dilute acetic acid) that was introduced during
synthesis at Cys21 and Cys38. When disulfide bond
formation was not desired, the Cys residues in the original
MGD1 sequence were replaced by Ser. Peptides were
labelled with biotin by elongation during the solid-phase
synthesis with the spacer motif Ser-Gly-Ser followed by
N-terminal biotinylation. Isoelectric points were computed
from the amino acid sequences using the internet tool,
/>Sequence and structural analysis
Sequences of the arthropod defensin family were extracted
from the Pfam database [28]. The N- and C-termini of the
sequences corresponding to the defensin structural domain
that were sometimes missing in the Pfam alignments were
added manually. Sequence analysis was performed using
CLUSTAL X

minimal inhibitory concentration (MIC) was evaluated by
testing serial doubling dilutions and defined as the lowest
peptide concentration that prevented any growth [29]. The
bactericidal capacity of peptides was assessed using the
Live/Dead Bac Light Bacterial viability kit (Molecular
Probes). The fluorescence given by live (FITC SYTO9Ò,
green) or dead bacteria (propidium iodide, red) was
observed using a fluorescent microscope (Leica) equipped
with Omega filters XF22 and XF 32.
Antifungal assay
Susceptibility of Fusarium oxysporum (a gift from A. Vey,
INRA Saint Christol-le
`
s-Ale
`
s, France) and Candida sp.
(a gift from O. Thaler, Universite
´
Montpellier 2, France)
was tested by a liquid growth inhibition assay as described
by Fehlbaum [30]. Briefly, 80 lL of fungal spores (final
concentration 10
4
sporesÆmL
)1
) suspended in potato dex-
trose broth (Difco) containing 0.1 mg tetracycline was
addedto20lL of peptide dilutions in microtiter plates.
Peptides were replaced by 20 lL of sterile water in
controls. Growth inhibition was observed under the

Cytotoxicity tests on the human lymphoma K562 cell line
Peptide concentrations corresponding to 10 times the MIC
for M. lysodeikticus were incubated with K562 cells. Toxi-
city was evaluated after 48 h of incubation by measuring the
optical density of the culture at 570/690 nm using the
In Vitro Toxicology Assay Kit (Sigma), based on conver-
sion of the yellow tetrazolium salt MTT into purple
formazan crystals by metabolically active cells.
Results
Anti-Gram-positive bacteria activity is conveyed
by the cyclized loop 3
Figure 1A shows the three-dimensional structure of
MGD1 [16] and Fig. 1B the designed set of peptides.
Peptides with two cysteines were oxidized so as to be
cyclised. Dilutions of the purified peptides were further
tested for growth inhibition of the Gram-positive bacteria
M. lysodeikticus (Fig. 1B). Peptides corresponding to the
a-helical part of MGD1 (peptide T) or to the a-helical
part prolonged by the N-terminal turn (GFGSP) and by
the short sequence (IPGR) connecting the a-helix to the
first strand of the b-hairpin (peptide S) did not exhibit
measurable activity, although the latter peptide repre-
sents almost 50% of the MGD1 amino acid sequence.
2806 B. Romestand et al.(Eur. J. Biochem. 270) Ó FEBS 2003
However, the 9-mer peptide B, CGGWHRLRC, corres-
ponding to the sequence of the MGD1 loop 3 occurring
between the two b-strands, had an MIC of 28 l
M
(i.e.
about 2.5% of the activity of the synthetic MGD1,

b-strands apparently did not convey activity by themselves
as peptide X was inactive and peptide K active
(MIC ¼ 12 l
M
). Peptide X corresponds to peptide K in
which the sequence of the b-strands was maintained but
the amino acids from loop 3 had been replaced by
multiple Ser and Gly residues. Therefore, in the sequence
of the whole b-hairpin structure of MGD1, only the loop
part seems to convey activity.
Activity is directed mainly against Gram-positive bacteria
and fungi
The activity spectra of peptides B, K, M and Q were
compared with that of synthetic MGD1 (peptide A).
Although less active than the entire molecule, peptides
derived from loop 3 were active on all the Gram-positive
bacteria tested (Table 1). Gram-negative bacteria were not
inhibited by any peptide, with the exception of E. coli 363,
which was sensitive to peptides K and M (MIC ¼ 62 l
M
),
and Q (MIC ¼ 22 l
M
). The fungus F. oxysporum was
inhibited by all four peptides, especially by peptides Q and
M(MIC¼ 13–15 l
M
) and peptide K (MIC ¼ 17 l
M
).

Lys (peptides C, D and J, respectively) had higher pI values
and displayed greater inhibitory activity than that of
peptide B. A strict quantitative relationship between the
theoretical isoelectric points and the logarithm of the
corresponding experimental bacteriostatic activities (corre-
lation coefficient of 0.999) was observed (Fig. 2). Finally,
the increase in bacteriostatic activity observed with loop
3-based peptides as a function of their increasing pI (cationic
charges) was also observed for the activity of larger peptides
containing the substituted peptide B (Table 2, L–K, N–M
and P–Q), indicating that the properties of the loop drive the
properties of larger peptides enclosing the loop.
Binding capacity of loop 3-derived peptides
on
M. lysodeikticus
The aforementioned results showed that synthetic peptides
corresponding to fragments of the MGD1 defensin
reproduced the behaviour of the entire molecule with
regard to its specificity. Thus, an active synthetic peptide
can be used instead of the natural molecule to study the
interaction of defensin with the bacterial membrane. To
monitor the mode of action on bacteria, biotin-labelled
peptides B and D were incubated with M. lysodeikticus
and binding of the peptides to bacteria was examined
using FITC-streptavidin. At concentrations of 1–60 l
M
,
both biotinylated peptide D (Fig. 3A) and biotinylated
peptide B (not illustrated) decorated the cell surface but
apparently did not penetrate the bacteria. Even using

on Gram-
positive bacteria M. lysodeikticus. The theoretical isoelectric point was
computed ( and plotted against
the log of the measured MIC.
2808 B. Romestand et al.(Eur. J. Biochem. 270) Ó FEBS 2003
status of M. lysodeikticus bacteria treated with 30 l
M
peptides E, B and D was assessed using a double
fluorescence labelling (Fig. 3B). In the absence of any
peptide or in the presence of peptide E (noncyclized
loop 3), an important number of live bacteria was
observed and practically no dead cells. However, both
peptides B and D inhibited the growth of bacteria, leading
to a low number of observable green fluorescence. In
addition, the few detectable bacteria were dead.
Common features of sequences of loop 3 in arthropod
and mussel defensins
Figure 4 shows the
CLUSTAL
format alignment of arthro-
pod defensins. Two subfamilies were identified by this
analysis, one including MGD1 (structural PDB code
1FJN) as a prototype and one including the insect
defensin A (PDB code 1ICA). In the MGD1 subfamily,
some striking features of the loop 3 sequences (comprised
between conserved cysteines Cys25 and Cys33 in MGD1)
are evident: (a) this part of the molecule contains at least
one (often two) positively charged residue (Lys or Arg;
boxed characters in Fig. 4); (b) it contains one or several
hydrophobic amino acids (Phe, Trp, Leu; greyed charac-

[20,21], the mode of action of defensins requires an initial
binding step on the outer membrane. The way this contact
takes place and the molecular features of the protein
involved are yet to be deciphered. The lack of information
about the mode of action and the availability of a refined
three-dimensional model [16] prompted us to prepare a
series of synthetic fragments designed on the basis of the
secondary structure elements of MGD1.
A remarkable result is that only peptides including
residues 25–33 of MGD1 displayed activity against Gram-
positive bacteria and fungi, after this short peptide had
been cyclized by disulfide bridging. Three series of
arguments suggest that the b-hairpin loop of MGD1,
i.e. residues 25–33 (CGGWHRLRC), plays a major role
in the binding of MGD1 to M. lysodeikticus. First, among
the synthetic fragments that we designed from the
available three-dimensional structure, only the cyclic
peptide CGGWHRLRC showed bacteriostatic activity
whereas larger fragments, corresponding either to the
a-helix sequence, or to the a-helix sequence extended by
the loop between the a-helix and the first b-strand (loop
2), or to the sequence of the whole b-hairpin with residues
from the loop substituted by serine and glycines, had no
detectable activity. This cyclic peptide CGGWHRLRC
was observed in confocal microscopy to bind to M. lyso-
deikticus, inhibiting the bacterial growth without lysing the
bacteria. Therefore, it is speculated that the binding of
MGD1 to the bacterial membrane is mediated by the
loop 3 region of the defensin, thus participating in the
early events of bactericidal activity. Our construction of

)1
of
biotinylated peptide D at 37 °C for 24 h and the binding of peptides to bacteria visualized with FITC–streptavidin. Confocal microscopic images
show the localization of the biotinylated peptide D on the cell surface. Similar results were obtained for peptide B. Control experiments were
performed in the presence of FITC–streptavidin and absence of peptide, and in the presence of FITC–streptavidin and 30 l
M
of an irrelevant
biotinylated peptide (Biot-YKKWINTFSGVPTYA). (B) Viability of M. lysodeikticus in the presence of 30 l
M
peptide E, B or D after overnight
incubation at 37 °C. The live or dead status of bacteria was assessed by labeling with FITC SYTO9Ò (green fluorescence, living bacteria) and
propidium iodide (red fluorescence, dead bacteria). Bacterial growth in the absence of any peptide was used as a control. Note the absence of killing
in the presence of peptide E and the important number of green living bacteria. In contrast, both peptides B and D inhibited bacterial growth and
the few observed bacteria were dead.
2810 B. Romestand et al.(Eur. J. Biochem. 270) Ó FEBS 2003
important for the mode of action. In addition, this loop
has been found to be highly solvent exposed in MGD1
[16], defensin A [19] and the Raphanus sativus defensin
[31]. It is worth noting that many other cysteine-rich
antibacterial peptides, not belonging to the defensin
family, exhibit a cysteine-bridged loop that also contain
basic and hydrophobic residues: e.g. CRIVVIRVC (bac-
tenecin), CYRGIGC (tachyplesin), CRRRFC (buthinin),
CTMIPIPRC (tigerinin), etc. Also remarkable is the
observation that lactoferricin B (a tryptophan/arginine
rich antibiotic peptide), when enzymatically cleaved from
lactoferrin adopts a twisted b-sheet structure, the loop
part of which includes one tryptophan and two arginine
residues [32], thus resembling loop 3 derived peptides.
Third, the relationship between the isoelectric point of the

residues in yellow. (A) View from the top of the molecule. Accessible
and positively electrically charged residues form a linear patch. (B) Side
view showing layered hydrophobic and surface accessible residues.
Ó FEBS 2003 Role of the hairpin loop in mussel defensin activity (Eur. J. Biochem. 270) 2811
From our observations, it is not possible to infer whether
the key role of the b-hairpin loop of MGD1 (loop 3) in
antimicrobial activity of MGD1 is of general value in the
mechanism of action of defensins. However, our results
could be compared with results obtained by other groups
who also point to the role of the connecting loop of the
b-hairpin of defensins. A combination of mutational
analysis [35] and structural analysis of the plant defensin
Rs-AFP1 [31] identified two subsites on this molecule
comprising Ôresidues in the protruding domain consisting
of the type VI b-turn and the first part of b-strand 3Õ (i.e. the
b-hairpin loop and part of the adjoining b-sheet) and
Ôresidues in the loop connecting b-strand and a-helix and
contiguous residues on the a-helix and the last part of
b-strand 3Õ [35]. Our observations were similar, although
obtained by a completely different approach. Note that we
did not succeed in demonstrating activity with synthetic
peptides corresponding to the second subsite of the
Rs-AFP1; this might be due to methodological differences.
Finally, two reports using synthetic peptides derived from
the amino acid sequence of rabbit [36] and plant defensins
[37] led to the conclusion that the whole b-hairpin could be
an important structural feature of the mode of action,
although the precise role of the loop was not elucidated.
In conclusion, our results indicated that residues 23–35 of
mussel defensin MGD1 play a key role in the binding to

7. Broekaert, W.F., Terras, F.R., Cammue, B.P. & Osborn, R.W.
(1995) Plant defensins: novel antimicrobial peptides as compo-
nents of the host defense system. Plant Physiol. 108, 1353–1358.
8. Segura, A., Moreno, M., Molina, A. & Garcia-Olmedo, F. (1998)
Novel defensin subfamily from spinach (Spinacia oleracea). FEBS
Lett. 435, 159–162.
9. Charlet,M.,Chernysh,S.,Philippe,H.,Hetru,C.,Hoffmann,J.A.
& Bulet, P. (1996) Innate immunity. Isolation of several cysteine-
rich antimicrobial peptides from the blood of a mollusc, Mytilus
edulis. J. Biol. Chem. 271, 21808–21813.
10. Hubert, F., Noel, T. & Roch, P. (1996) A member of the
arthropod defensin family from edible Mediterranean mussels
(Mytilus galloprovincialis). Eur. J. Biochem. 240, 302–306.
11. Ganz, T. & Lehrer, R.I. (1994) Defensins. Curr. Opin. Immunol. 6,
584–589.
12. Lehrer,R.I.&Ganz,T.(1999)Antimicrobialpeptidesinmam-
malian and insect host defence. Curr. Opin. Immunol. 11, 23–27.
13. Ganz,T.&Lehrer,R.I.(1999)Antibioticpeptidesfromhigher
eukaryotes: biology and applications. Mol. Med. Today 5,
292–297.
14. Tang, Y.Q., Yuan, J., Osapay, G., Osapay, K., Tran, D., Miller,
C.J., Ouellette, A.J. & Selsted, M.E. (1999) A cyclic antimicrobial
peptide produced in primate leukocytes by the ligation of two
truncated alpha-defensins. Science 286, 498–502.
15. Bonmatin, J.M., Bonnat, J.L., Gallet, X., Vovelle, F., Ptak, M.,
Reichhart, J.M., Hoffmann, J.A., Keppi, E., Legrain, M. &
Achstetter, T. (1992) Two-dimensional 1H NMR study of
recombinant insect defensin A in water: resonance assignments,
secondary structure and global folding. J. Biomol. NMR. 2,
235–256.

24. Yeaman, M.R., Bayer, A.S., Koo, S.P., Foss, W. & Sullam, P.M.
(1998) Platelet microbicidal proteins and neutrophil defensin dis-
rupt the Staphylococcus aureus cytoplasmic membrane by distinct
mechanisms of action. J. Clin. Invest. 101, 178–187.
25.Gausepohl,H.,Boulin,C.,Kraft,M.&Frank,R.W.(1992)
Automated multiple peptide synthesis. Pep. Res. 5, 315–320.
26. Laune, D., Molina, F., Ferrieres, G., Villard, S., Bes, C., Rieunier,
F., Chardes, T. & Granier, C. (2002) Application of the Spot
method to the identification of peptides and amino acids from the
2812 B. Romestand et al.(Eur. J. Biochem. 270) Ó FEBS 2003
antibody paratope that contribute to antigen binding. J. Immunol.
Methods 267, 53–70.
27. Tam, J.P., Wu, C., Liu, W. & Zhang, J.W. (1991) Disulfide bond
formation in peptides by dimethyl sulfoxide. Scope and applica-
tions. J. Am. Chem. Soc. 113, 6657–6662.
28. Bateman, A., Birney, E., Cerruti, L., Durbin, R., Etwiller, L.,
Eddy, S.R., Griffiths-Jones, S., Howe, K.L., Marshall, M. &
Sonnhammer, E.L. (2002) The Pfam protein families database.
Nucleic Acids Res. 30, 276–280.
29. Hancock, R.E., Falla, T. & Brown, M. (1995) Cationic bacteri-
cidal peptides. Adv. Microb. Physiol. 37, 135–175.
30. Fehlbaum, P., Bulet, P., Michaut, L., Lagueux, M., Broekaert,
W.F.,Hetru,C.&Hoffmann,J.A.(1994)Insectimmunity.Septic
injury of Drosophila induces the synthesis of a potent antifungal
peptide with sequence homology to plant antifungal peptides.
J. Biol. Chem. 269, 33159–33163.
31. Fant, F., Vranken, W., Broekaert, W. & Borremans, F. (1998)
Determination of the three-dimensional solution structure of
Raphanus sativus antifungal protein 1 by 1H NMR. J. Mol. Biol.
279, 257–270.