Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
J. H. Bowie, Department of Chemistry, The University of Adelaide, South Australia, 5005, Australia, Fax: + 61 88303 4358, Tel.: + 61 88303 5767, E-mail: email@example.com
Nineteen citropin peptides are present in the secretion from the granular dorsal glands of the Blue Mountains tree-frog Litoria citropa; 15 of these peptides are also present in the secretion from the submental gland. Two major peptides, citropin 1.1 (GLFDVIKKVASVIGGL-NH2), citropin 1.2 (GLFDIIKKVASVVGGL-NH2) and a minor peptide, citropin 1.3 (GLFDIIKKVASVIGGL-NH2) are wide-spectrum antibacterial peptides. The amphibian has an endoprotease which deactivates these membrane-active peptides by removing residues from the N-terminal end: loss of three residues gives the most abundant degradation products. The solution structure of the basic peptide citropin 1.1 has been determined by NMR spectroscopy [in a solvent mixture of trifluoroethanol/water (1 : 1)] to be an amphipathic α-helix with well-defined hydrophobic and hydrophilic regions. The additional four peptides produced by the dorsal glands are structurally related to the antibacterial citropin 1 peptides but contain three more residues at their C-terminus [e.g. citropin 1.1.3 (GLFDVIKKVASVIGLASP-OH)]. These peptides show minimal antibacterial activity; their role in the amphibian skin is not known.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Amphibians have rich chemical arsenals in the skin glands that form an integral part of their defence system, and also assist with the regulation of dermal physiological action [1,2]. In response to a variety of stimuli, host defence compounds are secreted from these specialized glands onto the dorsal surface and into the gut of the amphibian. Some of these compounds are antibacterial peptides, e.g. the bombinin, brevinin and magainin families of peptides [2–4]. The magainin peptides also show some anti-cancer activity . Some of the antibacterial peptides have attracted pharmaceutical interest .
The dermal secretions of some Australian tree-frogs (Litoria splendida, L. caerulea, L. gilleni, L. xanthomera, L. chloris and L. genimacula) collectively contain more than 30 antibacterial peptides [7–14]. The major wide-spectrum antibacterial agents from these tree-frogs belong to either the caerin or maculatin families of peptides. Particular examples are caerin 1.1 (one of 10 structurally related peptides from L. splendida, L. caerulea, L. gilleni, L. chloris and L. xanthomera) and maculatin 1.1 (from L. genimaculata). These peptides are, respectively, 25 and 21 amino acids in length: their sequences are related but maculatin 1.1 is missing four amino acid residues from the centre of a caerin 1-type sequence. The two peptides have the following sequences:
Caerin 1.1 GLLSVLGSVAKHVLPHVVPVIAEHL-NH2
Maculatin 1.1 GLFGVLAKVAAHVVPAIAEHF-NH2
The solution structures of both peptides have been investigated by NMR spectroscopy. In trifluoroethanol/water mixtures, caerin 1.1 adopts two well-defined helices (Leu2 to Lys11 and from Val17 to His24) separated by a hinge region of less-defined helicity and greater flexibility . In contrast to caerin 1.1, maculatin 1.1 does not have such marked flexibility in its central region. In this case, NMR spectra [measured (a) in trifluoroethanol/water and (b) with the peptide incorporated into an artificial phospholipid micelle] show that the amphipathic α-helix of maculatin 1.1 is distorted by Pro15: the central axis of the helix preceding Pro15 makes an angle of approximately 20° to the axis of the helix following Pro15 .
Amphipathic peptides of this type are membrane-active antibacterial agents. Interaction occurs at the membrane surface with the charged (normally basic) peptide adopting an α-helical conformation and attaching itself to charged sites (normally anionic) on the lipid bilayer. Two potential mechanisms have been proposed for interaction of these peptides with membranes, viz: aggregation of peptides is followed either by (a) penetration through the lipid bilayer via formation of a transmembrane helical bundle (the barrel-stave mechanism) or (b) formation of pores through the membrane as a result of the alignment of the peptides parallel to the membrane surface (the ‘carpet’ mechanism). As a result, disruption of normal membrane function occurs leading to lysis of the cell [17–22]. The magainin peptides (from the African clawed frog Xenopus laevis) are the best studied of such amphibian peptides. NMR studies show they form well-defined amphipathic α-helices in solution and when incorporated into an artificial phospholipid; the peptides are positively charged and interact readily with anionic phospholipids [6,18,21–23]. As caerin 1.1 and maculatin 1.1 are also membrane-active basic peptides, then perhaps they react in a similar manner to the magainins. Although the solution structures of caerin 1.1 and maculatin 1.1 are quite different, the antibacterial activities of each peptide are very similar [15,16]. If they are synthetically modified to convert them into molecules with better defined α-helical structures, i.e. removal of the hinge of caerin 1.1 by conversion of both Pro residues to Ala, or distortion of maculatin 1.1 by conversion of Pro15 to Ala15, the spectrum of antibacterial activities is significantly reduced in each case.
In this paper we describe the isolation, sequence determination, and activities of the citropin peptides from the Blue Mountain tree-frog of Australia, L. citropa. This frog is found near flowing, rocky streams in heavily forested areas from the coast to the Dividing Ranges in eastern Victoria and New South Wales. The length of the adult frog is in the range 4.5–5.6 cm, and the frog can be distinguished by its brown dorsum, green flash colours near the head, and by a brown tympanum [24,25]. This tree-frog is unusual in that it has two types of skin glands, the granular glands on the dorsal surface and the large submental gland on the throat. Other Australian tree-frogs that we have studied have no submental gland, thus it is of interest to determine the role of this gland, for example, does it contain host defence peptides like the dorsal glands of other amphibians, and if so, whether these peptides are similar in activity and structure to the caerin 1 and maculatin 1 peptides.
Preparation of skin secretions
L. citropa was held by the back legs, the skin moistened with deionized water, and the two types of glands (i.e. the granular dorsal glands situated on the back, and the submental gland situated on the throat) were stimulated independently by means of a bipolar electrode of 21G platinum attached to a Palmer Student Model electrical stimulator. The electrode was rubbed gently in a circular manner on the particular gland (under study) of the animal, using 10 V and a pulse duration of 3 ms . The resulting secretion was washed from the frog with deionized water (50 mL), the mixture diluted with an equal volume of methanol, centrifuged, filtered through a Millex HV filter unit (0.45 µm), and lyophilized. This procedure provided, on average, some 5 mg of solid material from the dorsal glands and 3 mg from the submental gland. This work conforms with the Code of Practice for the Care and Use of Animals for Scientific Purposes (1990) and the Prevention of Cruelty to Animals Act 1985, and was approved by the University of Adelaide Animal Ethics Committee.
HPLC Separation of the glandular secretion
HPLC separation was achieved using a VYDAC C18 HPLC column (5 µ, 300 A, 4.6 × 250 mm) (Separations Group, Hesperia, CA, USA) equilibrated with 10% acetonitrile/aqueous 0.1% trifluoroacetic acid. The lyophilized mixture (generally ≈ 1 mg) (see above) was dissolved in deionized water (50 µL), of which a 10-µL fraction was injected into the column. The elution profile was generated using a linear gradient produced by an ICI DP 800 Data Station controlling two LC1100 HPLC pumps, increasing from 10 to 75% acetonitrile over a period of 60 min at a flow rate of 1 mL·min−1. The eluant was monitored by ultraviolet absorbance at 214 nm using an ICI LC-1200 variable wavelength detector (ICI Australia, Melbourne, Australia). The HPLC traces of the antibacterial peptide regions of each glandular extract (Fig. 1) are complex: 17 fractions were investigated. Fractions were collected, concentrated and dried in vacuo for subsequent analysis. Further HPLC separation was required for some fractions in order to obtain pure components.
Sequence determination of peptides
Electrospray mass spectroscopy. Electrospray mass spectra were determined using a Finnigan LCQ ion trap mass spectrometer. Purified fractions from the HPLC separation were dissolved in methanol/water (1 : 1, v/v) and infused into the electrospray source at 8 µL·min−1. Electrospray conditions were as follows: needle potential 4.5 kV, tube lens 60 V, heated capillary 200 °C and 30 V, sheath gas flow 208 kPa. Mass spectra were acquired with the automatic gain control on, a maximum time of 400 ms, and averaging over three microscans. Molecular masses of peptides less than 2000 Da in mass were determined from (M + H)+ ions; those greater than 2000 Da in mass were determined from (M + 2H)2+ ions. All peptides were methylated to form methyl esters. The mass of the MH+ parent ion of the methyl ester minus that of the MH+ of the original peptide allows the determination of the number of CO2H and CONH2 groups in the molecule (viz CO2H &Ægr; CO2Me = 14 Da, whereas CONH2 &Ægr; CO2Me = 15 Da) .
Protein sequencing. Automated Edman sequencing was performed by a standard procedure  using an applied Biosystem 492 procise sequencer equipped with a 900-A data analysis module. The best results were obtained using a disc of immobilon film treated with bioprene in ethanol, onto which the peptide was absorbed from aqueous acetonitrile (90%). The disc was pierced several times with a razor blade to aid the flow of solvent.
Preparation of synthetic peptides
Ten natural citropin peptides and two modified citropin 1.1 peptides were synthesized (by Chiron Mimotopes, Clayton, Victoria, Australia) using l-amino acids via the standard N-α-Fmoc method (full details including protecting groups and deprotection have been reported recently ). Each synthetic citropin was shown to be identical with the natural citropin by both electrospray MS and HPLC. The sample of citropin 1.1 prepared for the NMR study was 91% pure (by HPLC and electrospray MS).
Antimicrobial testing on synthetic citropin peptides was carried out by the Microbiology Department of the Institute of Medical and Veterinary Science (Adelaide, Australia). The method used involved the measurement of inhibition zones (produced by the applied peptide) on a thin agarose plate containing the microorganisms under study. The procedures are standard . Activities are recorded as the minimum inhibitory concentration (MIC) of peptide per mL required to totally inhibit the growth of the named microorganism.
CD spectra were recorded on a Jobin-Yvon CD-6 spectrophotometer at room temperature. CD spectra of citropin 1.1, at a concentration of 31 µm, were acquired in solutions of TFE in water (0–50% by volume). Each spectrum represents the average of five scans, with the data smoothed over n = ± 5 data points.
NMR spectroscopy of citropin 1.1
NMR experiments were carried out on a solution of 8 mg of citropin 1.1 dissolved in a mixture of water (0.35 mL) and d3-TFE (0.35 mL), giving a final concentration of 7.1 mm, at a measured pH of 2.25.
NMR spectra were acquired on a Varian Inova-600 NMR spectrometer, at a 1H frequency of 600 MHz and 13C frequency of 150 MHz. All NMR experiments were performed at 25 °C. 1H NMR spectra were referenced to the methylene protons of residual TFE (3.918 p.p.m.). The 13C (F1) dimensions of the heteronuclear single-quantum coherence (HSQC) and heteronuclear multiple-bond correlation (HMBC) spectra were referenced to the 13CH2 (60.975 p.p.m.) and 13CF3 (125.9 p.p.m.) resonances of TFE, respectively.
Double-quantum-filtered correlation spectroscopy (DQF-COSY) ; total correlation spectroscopy (TOCSY) ; and nuclear Overhauser effect spectroscopy (NOESY) ; were all collected in the phase-sensitive mode using time proportional phase incrementation  in t1. For each of these experiments, 256 t1 increments were used. In the TOCSY and DQF-COSY experiments, 16 scans were time averaged for each increment, while 48 scans were averaged in the NOESY experiments. The free induction decay in t2 consisted of 2048 data points over a spectral width of 5564.1 Hz. The transmitter frequency was centred on the water resonance; conventional low power presaturation from the same frequency synthesizer was applied during a 1.5-s relaxation delay to suppress the large water signal in the TOCSY and NOESY spectra. Gradient methods for water suppression were used in the DQF-COSY spectrum . The TOCSY spectrum was acquired with the pulse sequence used by Griesinger et al. , which minimizes cross relaxation effects by employing an MLEV-17 spin-lock of 70 ms duration. NOESY spectra were acquired with mixing times of 80, 150 and 250 ms.
An HSQC experiment  was acquired to assign the α-13C resonances via correlations to their attached protons. The interpulse delay was set to 1/2JCH = 3.6 ms corresponding to JCH = 140 Hz. For this experiment, 256 t1 increments, each comprising 16 time averaged scans, were acquired over 2048 data points and 5564.1 Hz in the directly detected (1H, F2) dimension. In the 13C (F1) dimension a spectral width of 24 140 Hz was used. An HMBC spectrum  was collected to assign the carbonyl-13C resonances via correlations through two and three bonds to α, β and NH protons (with an interpulse delay, 1/2JCH = 62.5 ms for JCH = 7 Hz). In this experiment 512 t1 increments, each comprising 64 scans, were acquired over 4096 data points and 5564.1 Hz in the 1H (F2) dimension. The spectral width for the 13C (F1) dimension was 40 000 Hz.
All 2D NMR spectra were processed on a Sun Microsystems Ultra Sparc 1/170 workstation using vnmr software (v 6.1a). The data matrices were multiplied by a Gaussian function in both dimensions and then zero-filled to 2048 data points in F1 prior to Fourier transformation, except in the case of the HMBC experiment where zero-filling was carried out on 4096 data points in F1. Final processed 2D NMR matrices consisted of 2048 × 2048 or 4096 × 4096 real points.
Cross-peaks in the TOCSY and NOESY (mixing time = 150 ms) spectra were assigned using xeasy software (v 1.3.13) . The volumes for these cross-peaks were converted to distance constraints using the method of Xu et al. . For each symmetrical pair of cross-peaks, the peak of smaller volume was used. This procedure generated 230 distance constraints. Of these, 133 were intraresidue constraints, 50 were sequential (i, i + 1) constraints and 47 were medium range constraints (from two to four residues distant). Twenty-six of the distance constraints were ambiguous.
3JNHCαH values were measured from a high-resolution 1D 1H NMR spectrum acquired with 0.012 Hz per point digital resolution (128 000 data points). Dihedral angles were restrained as follows: 3JNHCαH < 5 Hz, φ = –60° ± 30°; 5 < 3JNHCαH < 6 Hz, φ = –60° ± 40°. Where 3JNHCαH > 6 Hz, φ angles were not constrained. The absence of resolved splitting for an NH resonance implies that its 3JNHCαH value is < 5 Hz, i.e. the particular amino acid is part of a helical conformation. Hence, the φ angles for these types of residues were restrained to −60° ± 30°. A total of five dihedral angle restraints were used for the structure calculations.
Structures were generated on a Sun Microsystems Sparc 1/170 workstation using xplor software (v 3.851) [40,41]. The restrained molecular dynamics (RMD) and dynamical simulated annealing (SA) protocol was used , including the use of floating stereospecific assignments . Sum-averaging was employed to account for the ambiguous constraints. The all hydrogen distance geometry (ALLHDG) force field (v 4.03) was employed for all calculations . Initially a family of 60 structures was generated with random φ and ψ dihedral angles. These were subjected to 6500 steps (19.5 ps) of high temperature dynamics at 2000 K. The Knoe and Krepel force constants were increased from 1000 to 5000 kcal·mol−1·nm−2 and 200–1000 kcal·mol−1·nm−4, respectively. This was followed by 2500 steps (7.5 ps) of cooling to 1000 K with Krepel increasing from 1000 to 40000 kcal·mol−1·nm−4 and the atomic radii decreased from 0.9 to 0.75 times those in the ALLHDG parameter set. The last step involved 1000 steps (3 ps) of cooling from 1000 K to 100 K. Final structures were subjected to 200 steps of conjugate gradient energy minimization. The 20 structures produced with the lowest potential energies were selected for analysis. 3D structures were displayed using insight ii software (v 95.0, MSI).
Isolation, sequence determination and antibacterial activity of the citropin peptides.
A single specimen of L. citropa was kept in captivity for several years and an electrical stimulation procedure  was used to obtain glandular secretions. This can be carried out on a monthly basis without harming the animal. Mild electrical stimulation of the skin in the vicinity of (a) the granular dorsal glands, and (b) the submental gland, gave two glandular secretions: each of these was subjected to HPLC separation; those regions of the HPLC traces containing the antibacterial peptides are shown in Fig. 1. Nineteen peptides were identified and characterized following HPLC separation. The amino acid sequences of the 19 characterized peptides are listed in Table 1. We have named these peptides citropins. Fifteen of the citropin peptides are produced by both the dorsal and submental glands, however, there are four additional peptides produced only by the dorsal glands.
Table 1. The amino acid sequences of the citropin peptides from Litoria citropa. The nominal mass, equal to the sum of the integral masses of the amino acids, is shown. The designation of fractions in HPLC separation traces are shown in Figs 3 and 4. D, dorsal (on the back); S, submental (on the throat).
There are three groups of citropins, designated citropins 1–3, respectively. Eight of the peptides of the citropin 1 group are post-translationally modified, in that they have C-terminal CONH2 functionality. In contrast, four citropins 1 and all of the citropins 2 and 3 have C-terminal CO2H groups. Citropins 1.1, 1.2 and 1.3 can be represented by Edmundson projections in which the α-helical conformations have well-defined hydrophilic and hydrophobic zones (see, e.g. Fig. 2A for citropin 1.1). This suggests the possibility that citropins 1.1, 1.2 and 1.3 may be amphipathic α-helical peptides exhibiting antibacterial activity.
The situation concerning the citropins 2 and 3 is not as straightforward, however. Both groups of peptides have C-terminal CO2H groups, a function which generally results in reduction in antibacterial activity of amphibian peptides [1,2,45,46]. The Edmundson projections of the citropins 2 and 3 do not show such well-defined hydrophilic and hydrophobic zones as do those of the citropins 1 (see Fig. 2B for citropin 2.1.3). However, such peptides will not be able to form ideal α-helices because the presence of central Gly and/or Pro residues have a destabilizing influence on helical formation . For example, citropin 2.1.3 contains Gly4, 7, 11 and 12 together with Pro20. Therefore, it is difficult to predict whether citropins 2 and 3 will show antibacterial or other bio-activity.
The 10 natural citropins shown in Table 2 were synthesized, their antimicrobial activities determined against the bacteria listed in Table 2, and the corresponding activities of caerin 1.1 and maculatin 1.1 (and uperin 3.6, see later) are included for comparison. Citropins 1.1, 1.2 and 1.3 show antibacterial activity against the same pathogens affected by caerin 1.1 and maculatin 1.1 (except that they show no activity against Pasteurella multocida, see Table 2). The wide-spectrum antibacterial activity of the three citropins 1 is more pronounced against Gram-positive than Gram-negative organisms. Neither citropin 2 nor citropin 3 peptides show significant activity against bacteria (Table 2).
Table 2. Antibacterial activities of some natural and modified citropin peptides from litoria citropa. Comparison with the antibacterial activities of caerin 1.1, maculatin 1.1 and uperin 3.6. MIC, minimum concentration of the peptide in µg·mL−1 necessary to inhibit the growth of the pathogen; sm, syn mod.
a If there is no figure listed the MIC value is > 100 µg·mL–1.
Two synthetic modifications of citropin 1.1 (Ala4 and Ala7 citropin 1.1) were prepared in order to determine the importance of the hydrophilic Asp and Lys residues on the antibacterial activity of citropin 1.1. These data are also listed in Table 2. The residue Asp4 is not essential for antibacterial activity: indeed, replacement of Asp4 with Ala4 marginally increases the antibacterial activity. In contrast, replacement of Lys7 with Ala7 significantly reduces the antibacterial activity of the peptide.
The solution structure of citropin 1.1
The interesting features of citropins 1.1, 1.2 and 1.3 are that they are very simple basic peptides that, with only 16 amino acid residues, are not long enough (when in an α-helical conformation) to traverse the lipid bilayer of a bacterial cell (e.g. via the ‘barrel stave’ mechanism: a minimum of 20 residues is needed for such a mechanism [19–21]). A prerequisite for antibacterial membrane activity may involve some aggregation of citropin 1 molecules on the membrane surface before the bacterial cell wall can be breached, e.g. via the so called ‘carpet’ mechanisms for the disruption of membrane function [19,20]. Do the citropins 1 adopt well-defined α-helical and amphipathic structures as predicted by Edmundson projections (cf. Fig. 2A), or do they have some other structure? We have used CD and NMR spectroscopy to investigate the solution structure of citropin 1.1.
CD spectra were acquired on citropin 1.1 in increasing concentrations (0–50%, v/v) of TFE in water. From 0 to 20% TFE, the CD spectra of citropin 1.1 showed a broad minimum at around 200 nm characteristic of an unstructured or random coil peptide. At concentrations of 30–50% TFE, the CD spectra exhibited two minima in the vicinity of 208 and 222 nm, indicating that citropin 1.1 adopts a predominantly α-helical structure under these solvent conditions. The greatest ellipticity at these minima occurred at 50% TFE suggesting maximal helical conformation at this concentration. Peptides with pronounced helical propensity reach a maximum helical content at concentrations such as this (i.e. 30–50% TFE) , suggesting citropin 1.1 has a high tendency to form a helical structure. Based on this evidence, NMR studies were performed using a 50% solution of the peptide in TFE/H2O.
The 1H NMR spectrum was assigned using xeasy software (v 1.3.13)  following the sequential assignment procedure , which involved the combined use of DQF-COSY, TOCSY and NOESY spectra. An HSQC spectrum was used to assign the α-13C resonances from the one-bond correlations to the assigned 1H resonances. Similarly, an HMBC spectrum was employed to make the carbonyl-13C assignments from the two- and three-bond correlations to the assigned 1H resonances. Table 3 lists all the assignments for the 1H, α-13C and carbonyl-13C resonances.
Table 3. 1H and 13C NMR chemical shifts for citropin 1.1 in TFE/H2O (1 : 1, v/v), at a measured pH of 2.25 at 25 °C. Assignments for all the 1H NMR resonances are presented whereas only the α-13C and carbonyl-13C resonances are tabulated; NO, not observed.
The NH region of a NOESY spectrum in which a series of sequential NH to NH NOEs [dNN(i, i + 1)] occur along the length of the peptide is shown in Fig. 3. A series of weaker dNN(i, i + 2) NOEs are also observed at a lower contour level in this region. A summary of the various types of NOEs observed for citropin 1.1 is presented in Fig. 4. Here it can be seen that in addition to the NOEs mentioned above, a number of weak sequential dαN(i, i + 1) NOEs occur as well as a series of NOEs from residues three and four amino acids apart [dαN(i, i + 3), dαβ(i, i + 3) and dαN(i, i + 4)]. Taken together, the observed NOEs and their intensities are consistent with citropin 1.1 having a helical structure along the majority of its sequence.
A helical structure for citropin 1.1 is also indicated from an examination of the deviation from random coil chemical shift values of the α-1H, α-13C and carbonyl-13C resonances determined in water [48,49]. Smoothed over a window of n = ± 2 residues , the plot for the 1H α-CH resonances shows a distinct upfield shift (Fig. 5), while those for the 13C resonances show a distinct downfield shift, except in the vicinity of the N- and C-termini. The directions of these deviations from random coil chemical shift values are consistent with the peptide having a helical structure along its length, with maximal helicity in its central region and less well-defined structure at its N- and C-termini [51–53].
The conclusions derived from an examination of the NMR data were confirmed when the NOE data were used as input for structural calculations. Sixty structures were generated by RMD and dynamical SA calculations and the 20 structures of lowest potential energy were selected for close examination. Some statistics for the 20 final structures are given in Table 4.
Table 4. Structural statistics of citropin 1.1 following RMD/SA calculations. < SA > is the ensemble of the 20 final structures (SA) is the mean structure obtained by best-fitting and averaging the coordinates of backbone N, α-C and carbonyl-C atoms of the 20 final structures. (SA)r is the representative structure obtained after restrained energy minimization of the mean structure. Well-defined residues are those with angular order parameters (S) > 0.9. For citropin 1.1, residues Leu2 to Gly15 are well-defined.
RMSD from mean geometry (nm)
< SA >
All heavy atoms
0.087 ± 0.016
All backbone atoms (N, α-C, carbonyl-C)
0.056 ± 0.016
Heavy atoms of well-defined residues
0.072 ± 0.013
Backbone atoms (N, α-C, carbonyl-C) of well-defined residues
0.037 ± 0.014
xplor energies (kcal·mol–1)
50.62 ± 2.37
2.98 ± 0.18
15.60 ± 1.01
1.90 ± 0.42
6.75 ± 1.03
23.39 ± 1.48
0.00 ± 0.01
The superimposition of the 20 structures over the backbone N, α-C and carbonyl-C atoms shows that citropin 1.1 forms a regular α-helix along its entire length (Fig. 6a). Analysis of the angular order parameters (S, φ and ψ)  of these structures indicated that, except for the N- and C-terminal residues (Gly 1 and Leu 16), all residues were well defined with S > 0.9 for both their φ and ψ angles (not shown). The (φ,ψ) distribution (i.e. Ramachandran plot ) for the well-defined amino acids were all in the favoured region for α-helical conformation (not shown). The most energetically stable of the final 20 structures is displayed in Fig. 6b and from this representation it is apparent that citropin 1.1 forms an amphipathic α-helical structure with well-defined hydrophobic and hydrophilic faces.
Comparison of the observed NH chemical shifts of citropin 1.1 with the corresponding random coil NH chemical shifts  revealed a periodic distribution such that those from residues in the centre of the hybophobic face of the amphipathic face were shifted upfield with respect to the random coil values. This behaviour was first noted and investigated by Zhou et al.  in their CD and NMR study of a synthetic amphipathic helical peptide. They concluded that these variations in NH chemical shifts arise from a slight curvature of the amphipathic helix, which leads to differences in backbone hydrogen bond length along both faces of the peptide. In the case of citropin 1.1, it is apparent from Fig. 6b that the helix is slightly curved such that the hydrophobic face is concave, i.e. it has shorter hydrogen bonds than the hydrophilic side.
Citropins 1.1 and 1.2 are the major antibacterial and amphipathic peptides produced by both the dorsal and submental glands of L. citropa. These peptides are structurally dissimilar to the antimicrobial peptides of other tree-frogs of the genus Litoria that we have studied previously both in sequences and in length . They are the simplest, amphibian, wide-spectrum antibacterial peptides reported to this time. They differ only in the nature of the hydrophobic residues at positions 5, 6, 12 and 13 in their sequence, and all contain the hydrophilic residues Asp4, Lys7 and 8 and Ser11. There is however, one feature of L. citropa that is similar to other tree-frogs of the genus Litoria that we have studied, i.e. they all produce endoproteases that deactivate the major antibacterial peptides after they have been on the skin for some period (after about 10 min for L. citropa: within the 5–30 minute range for other frogs we have studied [13,14]). In the case of caerin 1.1 and maculatin 1.1, endoproteases remove the first two amino acid residues effectively deactivating each peptide [13,14], while citropins 1.1 and 1.2 are principally degraded by the removal of three residues, with peptides with two missing residues being formed in much smaller yield (see Tables 1 and 2 for sequence and activity data, respectively) (cf. also the degradation of the magainin peptides from X. laevis). The citropin peptides show little sequence similarity to other antibacterial peptides from frogs of the genus Litoria so far studied, but they do show an overall structural resemblance to the 17 residue uperin 3 antibacterial peptides that are produced by toadlets of the genus Uperoleia. For example, compare the sequences of uperin 3.6 (a major antibacterial peptide of Uperoleia mjobergii) and citropin 1.1.
Citropin 1.1 GLFDVIKKVASVIGGL-NH2
These two peptides have a very similar distribution of hydrophobic and hydrophilic amino acids, in particular the three conserved charged amino acids which are highlighted in bold. In addition to these amino acids, uperin 3.6 contains an additional Lys at residue 14. The spectrum of antibacterial activity of uperin 3.6 is very similar to those of citropin 1.1, 1.2 and 1.3 (see Table 2). The solution structure of uperin 3.6 has been investigated by NMR techniques , and like citropin 1.1, it is a well defined amphipathic α-helix along the entire length.
Although citropin 1.1 is almost entirely helical in the membrane-mimicking solvent TFE (Fig. 6), the length (16 amino acids) is too short to span the lipid bilayer. The potent antibacterial activity of citropin 1.1 cannot therefore arise from simple aggregation of the peptides in an orientation perpendicular to the bilayer plane to form a pore (e.g. via the barrel-stave mechanism). Other mechanisms for the lytic activity of the antibacterial citropin and uperin peptides must therefore be invoked, e.g. those involving the orientation of the peptides parallel to the membrane surface where they either aggregate in a carpet like manner  or diffuse continuously throughout the membrane . In both instances, destabilization of the membrane occurs which creates openings within the bilayer.
What are the roles of the citropin 2 and 3 peptides in the amphibian integument? The complex citropin 2 and 3 peptides (Table 1) have not shown significant antibacterial activity in our testing regime. The role of the basic citropin 2 peptides is not known at this time. The citropin 3 peptides are quite unusual in that (a) the sequence commences with Asp (the only other reported amphibian peptides that commence with Asp are hylambin  and kassinin ; both members of the tachykinin family of neuropeptides which bear no structural relationship to the citropins 2 and (b) they are anionic peptides. The anionic nature of the peptides suggests that they may be spacer peptides for the procitropin 1 peptides within the dorsal and submental glands (see  for a review of the biosynthesis of antibiotic peptides), although this proposal has not been confirmed.
Finally, the role of the four extra citropin 1 peptides from the dorsal gland secretion (I to L in Fig. 2 and Table 1), is, to date, a mystery. The four extra citropin 1 peptides are related structurally, two to citropin 1.1 and two to citropin 1.2. The last two residues of the citropins 1.1 and 1.2 have been replaced by Leu-Asp and two extra residues (either Ser-Pro-OH or Ser-Gln-OH) added. One of these peptides, citropin 1.1.3, has been tested for antibacterial activity, but such activity is insignificant (see Table 2). Perhaps these peptides are hormones or neuropeptides. If so, they bear no structural resemblance to any known amphibian hormone or neuropeptide: they are certainly not related to the potent amphibian caerulein or uperolein type neuropeptide vasodilators [1,2,45,46].
This work was financed by the Australian Research Council. K.L.W., P.A.W. and B.C.S.C. acknowledge the awards of an Australian Postgraduate Award, an ARC PhD scholarship, and an Overseas Australian Postgraduate Award, respectively. We thank Mr G. Beardsell for providing access to a specimen of L. citropa, Ms D. Turner for automated peptide sequencing, and Dr R. Lindner for assistance with acquisition and processing of the CD spectra.
Fig. S1. CD spectra of citropin 1.1 at the varying concentrations (v/v) of trifluoroethanol (TFE) indicated.
Fig. S2. Deviation from random coil chemical shifts  of (a) α-13C and (b) carbonyl 13C resonances of citropin 1.1 in 50% (v/v) d3-TFE. A positive chemical shift difference indicates a downfield chemical shift compared to the random coil value. The deviation values were smoothed over a window of n = ± 2 residues .