Structure–function relationships of temporins, small antimicrobialpeptides from amphibian skin


D. Barra, Dipartimento di Scienze Biochimiche, Università‘La Sapienza’, Piazzale Aldo Moro 5, 00185 Roma, Italy. Fax: + 39 06 4440062, E-mail:


Temporins, antimicrobial peptides of 10–13 residues, were isolated from secretions of Rana temporaria[Simmaco, M., Mignogna, G., Canofeni, S., Miele, R., Mangoni, M.L. & Barra, D. (1996) Eur. J. Biochem.242, 788–792]. These molecules are specific to this amphibian species, which is also able to secrete on its skin other antimicrobial peptides similar to those found in different Rana species. The effect of temporins A, B and D (13 residues, net charge +2), and H (10 residues, net charge +1 and +2, respectively) against both artificial membranes of differing lipid composition and bacteria has been investigated in order to gain insight into their mechanisms of action. The results indicate that: the lytic activity of temporins is not greatly affected by the membrane composition; temporins A and B allow the leakage of large-size molecules from the bacterial cells; temporin H renders both the outer and inner membrane of bacteria permeable to hydrophobic substances of low molecular mass; and temporin D, although devoid of antibacterial activity, has a cytotoxic effect on erythrocytes. The results allow important conclusions to be drawn about the minimal structural requirements for lytic efficiency and specificity of temporins.


colony-forming unit


2-nitrophenyl-β-d galactoside




lethal concentration









Skin secretions of several amphibian species have been analysed in detail and found to contain a large number of different antimicrobial peptides, which represent the effector molecules of innate immunity [1,2]. Recent experiments on Rana esculenta or Bombina orientalis have also demonstrated that in amphibia genes for antimicrobial peptides are controlled by NF-κB-regulated transcription [3–6].

All anti-microbial peptides studied so far have a cationic character that allows their preferential interaction with the anionic phospholipids of the target bacterial membranes. Some of them have been shown to lyse, with a certain extent of selectivity, cancer cells, which also contain negatively charged phospholipids, whereas normal cells do not [7]. Others are able to interact with erythrocytes causing a rapid hemolysis. Most of the antimicrobial peptides can adopt an amphipathic α-helical structure in hydrophobic environments, thus perturbing the phospholipid bilayer of the target membrane [8,9]. Inhibition of growth as well as cell death may then be a consequence of the disturbance of membrane functions.

A large variety of peptides of 20–46 amino acid residues has been isolated from frogs belonging to the Rana genus and found to be active against bacteria and fungi. These are the brevinins, esculentins and their related forms, ranalexin, gaegurins and rugosins: all these peptides share a peculiar structural motif, i.e. an intramolecular disulfide bridge forming a heptapeptide ring at the C-terminal end [1]. From Rana temporaria, a family of small (10–13 residues) antibacterial peptides called the temporins have also been isolated; so far they have been detected only in this amphibian species [10].

Temporins are among the smallest antimicrobial peptides so far described, together with the 13-residue peptide indolicidin [11] and the cyclic dodecapeptide bactenecin [12], both from bovine neutrophils. The last two are highly hydrophobic cationic peptides with a net charge of +4. Bactenecin contains a disulfide bridge, which confers a looped structure on the peptide, while indolicidin is amidated at the C-terminus. Both peptides are mainly active against Gram-negative bacteria [13,14]. Temporins are all amidated at the C-terminus; those containing one basic residue, either lysine or arginine, in the sequence (net charge +2) were found to be active specifically against Gram-positive bacteria and Candida albicans. When assayed against laboratory bacterial strains, temporins lacking a basic residue were inactive, and the 10-residue members of this family, although containing a basic residue, were inactive also [10].

To gain furher insight into the mechanism of action of these small antimicrobial peptides, we have investigated their effects against both artificial membranes of differing lipid composition and bacteria.

Materials and methods


Egg yolk l-α-phosphatidylcholine (PtdCho), bovine brain l-α-phosphatidyl-l-serine (PtdSer), l-α-phosphatidyl-dl-glycerol (PtdGro) enzymatically converted from PtdCho, and calcein were obtained from Sigma.

Synthetic temporins were purchased from Tana Laboratories (Houston, Texas). The purity of the peptides was checked by HPLC analysis, and their sequences were assessed by both automated Edman degradation with a Perkin-Elmer 476A protein sequencer and mass spectrometry with a Finnigan LCQ instrument. Peptide concentrations were determined by quantitative amino acid analysis with a Beckman System Gold instrument equipped with an ion-exchange column and ninhydrin derivatization.

CD measurements

CD spectra were obtained on a Jasco J710 spectropolarimeter, equipped with a DP 520 processor, at 20 °C, using a quartz cell of 2-mm pathlength. Spectra were the average of a series of 3–7 scans. Ellipticity is reported as the mean molar residue ellipticity (θ) expressed in deg·cm2·dmol−1. Peptide concentrations were determined by quantitative amino acid analysis.

Preparation of vesicles and leakage measurement

A 60-mm calcein solution (pH 7.0, 1 mL) was cosonicated with a lipid solution containing either PtdCho, PtdSer or PtdGro in chloroform (1 mL). Lipid vesicles were prepared through the reverse phase evaporation method [15]. Untrapped calcein was removed by gel filtration (Sephadex G-50, 1.5 × 15 cm column, equilibrated and eluted with 50 mm sodium phosphate buffer, pH 7.4, containing 0.1 mm EDTA). The lipid concentration in the separated vesicular fractions was determined by the method of Stewart [16]. Aliquots of a temporin solution in 20% EtOH were injected in a cuvette containing the stirred vesicular suspension in 50 mm sodium phosphate buffer, 0.1 mm EDTA, pH 7.4. The release of calcein from vesicles was monitored fluorometrically at 517 nm, exciting at 490 nm, on a Perkin-Elmer LS 50 B spectrofluorometer. The calcein that is entrapped in the vesicles is at high concentration and the fluorescence is self-quenched [17]. Leakage was monitored as a relief of quenching; the fluorescence intensity corresponding to 100% leakage was determined by addition of Triton X-100 (10%, v/v, in water) to the sample, which caused the total destruction of the vesicles with consequent total relief of the quenching. All experiments were conducted at room temperature.


The standard bacterial strains were Escherichia coli D21, D21f1, D21f2, D21e7, D22, Staphylococcus aureus Cowan I. The natural bacterial strains were: Enterobacter agglomerans Bo-1, Aeromonas hydrophila Bo-3 and Bo-4, Acinetobacter junii Bo-2, isolated from Bombina orientalis;Klebsiella pneumoniae Rt-1, Proteus vulgaris Rt-3 and A. junii Rt-4 isolated from R. temporaria[4].

Anti-bacterial assays

The antibacterial activity was evaluated using an inhibition zone assay on Luria–Bertani (LB)-agarose plates [18]. To study the bactericidal effect and the rate of killing of temporins, the peptides were added to a bacterial suspension and the number of surviving bacteria was followed at different incubation times. E. coli and S. aureus were grown at 30 °C in LB medium to an approximate D590 nm of 1 and then diluted in LB to about 1 × 104 or 6 × 104 cells per mL, respectively. Temporin A was added to E. coli diluted culture at a final concentration of 160 or 250 µm. Temporins A and B were added to S. aureus culture at 60 µm final concentration. Aliquots of 10 µL were withdrawn at different intervals and spread on agar plates. After overnight incubation at 30 °C, the surviving bacteria, expressed as colony-forming units (c.f.u.), were counted.

The synergistic effect of temporin H (300 µm) with rifampicin (1.5 µg·mL−1) against E. coli D21 was also studied either in LB medium or 0.9% NaCl.

The hemolytic activity was measured on human red cells both in solid medium, using a modification of the antibacterial inhibition zone assay, and in liquid medium as reported [19]. Serial dilution of the peptides were used, and after 30 min at 30 °C the cells were centrifuged and the absorbance in the supernatant was recorded at 415 nm. Complete lysis was measured by suspending erythrocytes in distilled water.

Cytoplasmic membrane permeabilization

Inner membrane pemeability was determined by measuring the β-galactosidase activity using 2-nitrophenyl-β-d galactoside (Gal-ONp, Sigma) as substrate according to [20]. The bacterial strain E. coli D22, which has a permeable outer membrane due to a mutation in the env A gene [21], was used. Cells were grown at 37 °C in LB medium supplemented with 1 mm isopropyl thio-β-d-galactoside to an approximate D590 nm of 1, washed and resuspended in 10 mm sodium phosphate buffer, pH 7.4. About 3 × 106 cells were incubated for 60 min at 30 °C with 100 µm of each peptide. The bacterial culture was then passed through a 0.2 µm filter and the hydrolysis of Gal-ONp was recorded on the filtrate at 420 nm.

Alternatively, for temporins D and H, the β-galactosidase activity on Gal-ONp was measured on the whole bacterial culture in the presence of the test peptide (100 µm).


Features of temporins

The sequences of the temporins tested in this study are reported in Table 1, as well as sequences of other antimicrobial peptides mentioned in this paper. These are magainin-2 and melittin, for which many structural and functional data are available [22–24]; brevinin-2T and a bombinin-like peptide (BLP-1) from amphibian skin [1]; bactenecin [12], crabrolin [25], indolicidin [26] and SPF [27,28], whose length is similar to that of temporins.

Table 1. Sequences of selected antimicrobial peptides.
Temporin HLSP–––NLLKSLL-amide

Mean residue hydrophobicities (H) and hydrophobic moments (µ) per residue were calculated using the Eisenberg consensus scale of hydrophobicity [29]. For temporins, ranges of H = 0.2308–0.0320, and µ = 0.360–0.258 were found. In both measures, temporin H has the lowest values.

CD measurements of temporins

CD spectra of synthetic temporins were recorded in water and after addition of trifluoroethanol. The experiments demonstrated that an increase in trifluoroethanol concentration caused a progressive change from a random conformation to an α-helical structure, the effect being complete at about 30% trifluoroethanol (Fig. 1). Interestingly, a very similar behaviour was observed for the decapeptide temporin H, with only slight differences in the shape of the spectra and dependence on trifluoroethanol concentration.

Figure 1.

CD spectra of temporins A and H (20 µm). CD spectra of temporins B and D are identical to those of temporin A. The helical wheel projections for each peptide are also shown.

Permeabilization of lipid vesicles

Dose–response curves for temporins A, B, D and H with PtdCho and PtdSer liposomes are shown in Fig. 2. In the experimental conditions used, all tested temporins led to the leakage of entrapped calcein from both types of lipid vesicles, although large differences in the relative efficiency were observed. In all cases, temporins A and B showed the greatest membrane-permeabilizing activity, and caused an almost total disruption of vesicles at a peptide concentration of 4.6 µm.

Figure 2.

Dependence of calcein leakage from PtdCho (A) or PtdSer (B) liposomes on peptide concentration. Calcein leakage is defined as the percent leakage after 10 min at a lipid concentration of 52–54 µm in 50 mm sodium phosphate buffer, 0.1 mm EDTA, pH 7.4. All experiments were performed at room temperature. ○, temporin A; ●, temporin B; □, temporin D; ▪, temporin H.

From the data presented in Fig. 2, a determinant effect of lipid composition on temporin-induced lysis cannot be detected. Liposomes consisting of zwitterionic (electrically neutral) phospholipids (PtdCho) are lysed as well as liposomes made of acidic phospholipids (PtdSer). To further explore whether surface charge has a role in conferring membrane selectivity to temporins A and B, L/P50 values, i.e. the lipid to peptide molar ratios at which 50% leakage is observed after 5 min incubation, with neutral (PtdCho) and negatively charged (PtdSer and PtdGro) lipid vesicles were measured (Table 2). It is evident that the presence of acidic phosholipids is not an absolute requirement for the temporin-induced lysis of biomembranes, although an increasing effect of PtdGro on the permeabilizing activity of temporin B can be noticed.

Table 2. Effects of lipid composition on liposome lysis induced by temporins A and B. (L/P)50, the lipid to peptide ratio at which 50% lysis is observed 5 min after addition of the peptide, is listed as a function of lipid composition. A larger value implies a stronger activity. Values are the mean of two independent measures. Other experimental conditions as in Fig. 2.
Lipid compositionTemporin ATemporin B

The time course of the release of calcein from PtdCho liposomes induced by temporins A and B is shown in Fig. 3. The leakage process consists of an initial rapid phase, followed by a slow phase. Moreover, for both temporins A and B, the initial rate of leakage, measured as percent leakage after 1 min incubation, was approximately linearly dependent on peptide concentration (Fig. 3C). The true initial rate is likely to be too rapid to be measured precisely without a stopped-flow apparatus. At the highest peptide concentration tested (4.6 µm), more than 90% of the total leakage occurred during the first minute after addition of the peptide. Similar results were also obtained for the action of temporins A and B on PtdSer and PtdGro liposomes (data not shown).

Figure 3.

Kinetics of temporin-induced lysis of PtdCho liposomes at various peptide concentrations: 4.6 µm(○); 2.3 µm(●); 1.16 µm(□); 0.58 µm(▪); 0.29 µm (▵). (A) Temporin A, (B) temporin B. Lipid concentration and other experimental conditions were as in Fig. 2. (C) Initial calcein leakage rate, expressed as percent leakage after 1 min, as a function of peptide concentration; ●, temporin A; ▴, temporin B. Data are taken from (A) and (B).

Anti-bacterial activity

The bactericidal activity and the rate of killing by temporins were studied against E. coli D21 and S. aureus Cowan I. According to the lethal concentration (LC) values obtained by the inhibition zone assay [10], in LB medium temporins D and H were completely inactive, whereas temporins A and B showed a higher activity against the Gram-positive bacterial strain than against the Gram-negative one. In fact, against S. aureus, temporins A and B at 60 µm concentration caused about 100% reduction in colony-forming units within 15 min, temporin A acting faster than temporin B (Fig. 4). In contrast, a four-fold higher concentration of temporin A was needed to kill, in 15 min, the same amount of E. coli cells (Fig. 5).

Figure 4.

Rate of killing of S. aureus Cowan I by temporins A and B at 60 µmin LB medium. The number of surviving bacteria is expressed as c.f.u. per mL. The control is given as bacteria without peptide.

Figure 5.

Rate of killing of E. coli D21 by temporin A at 160 and 250 µmin LB medium. The number of surviving bacteria is expressed as c.f.u. per mL. The control is given as bacteria without peptide.

Temporins A and B were also tested by the inhibition zone assay against cell-wall defective mutant strains of E. coli D21 (D21e7, D21f1 and D21f2), which have lost increasing amounts of sugar residues in their lipopolysaccharide (LPS) [30], and against E. coli D22 [21]. The square diameters of the bacteria-free zones obtained at different peptide concentrations are reported in Fig. 6. The increasing sensitivity of the envelope mutants to both peptides parallels their carbohydrate moiety reduction. Reduction or absence of sugar residues might facilitate the interaction of the peptide with the negatively charged lipid A.

Figure 6.

Anti-bacterial activity of different concentrations of temporin A (A) and temporin B (B) on E. coli D21, D22 and LPS mutants of E. coli D21. The activity is given as square diameters of the inhibition zones [18] obtained by depositing 3 µL of serial dilutions of the peptide solution in a 3 mm diameter well.

The antimicrobial activity of temporins was also studied against bacterial strains isolated from the natural flora of the frog [4]. The LC values are given in Table 3, in which brevinin 2T and bombinin BLP-1 are reported as reference peptides. Temporins D and H were inactive against all the strains tested. Although most of the bacteria displayed a similar susceptibility to both temporins A and B, against A. junii Bo-2 temporin B is about 10-fold more potent than temporin A. This suggests that these molecules have a differential effect against bacteria of the natural flora.

Table 3. Antimicrobial activity of amphibian peptides on frog natural bacterial strains.
 Lethal concentration (µM)
Bacterial strainBrevinin-2TTemporin ATemporin BBombinin (BLP-1)
Klebsiella pneumoniae Rt-10.59.610.07.4
Proteus vulgaris Rt-37.5> 340> 360> 180
Acinetobacter junii Rt-
Acinetobacter junii Bo-28.524.02.31.3
Aeromonas hydrophila Bo-330.0> 360> 36026.0
Aeromonas hydrophila Bo-46.5> 20> 3014.0
Enterobacter agglomerans Bo-12.1> 50> 501.9

In order to investigate the capacity of temporins in permeabilizing the bacterial outer membrane, the synergistic effect of the peptides with rifampicin was studied against E. coli D21, using the inhibition zone assay: 6 nmol of the test peptide were put in the well and the inhibition zone was compared with the one produced in agarose plates containing a sublethal concentration of rifampicin (2 µg·mL−1). Whereas no difference was found with temporins A, B and D, the antimicrobial activity of temporin H in the presence of rifampicin was 10-fold higher than temporin H alone (Fig. 7), suggesting that this peptide is able to facilitate the entrance of rifampicin into the bacterial cell, where it exerts its toxic effect.

Figure 7.

Anti-bacterial activity of temporins A, B and H (6 nmol) on E. coli D21 with and without rifampicin (2 µg·mL−1). Temporin D is not reported in the figure, being inactive in both cases. The activity is given as square diameters of the inhibition zones [18].

The synergistic action of temporin H and rifampicin was also studied in liquid medium, using either LB or 0.9% NaCl. The susceptibility of E. coli D21 to temporin H in the presence or absence of rifampicin was determined as c.f.u. during a 60 min incubation at 30 °C. The results reported in Fig. 8 show that in both LB medium or 0.9% NaCl the synergistic effect is remarkable.

Figure 8.

Rate of killing of E. coli D21 by temporin H (300 µm) in LB medium or 0.9% NaCl, with and without 1.5 µg·mL−1 rifampicin. The number of surviving bacteria is given as percentage of c.f.u.

In order to assess the properties of temporins in permeabilizing the bacterial inner membrane, the β-galactosidase activity in the supernatant of an E. coli D22 culture was measured using the chromogenic substrate Gal-ONp. The supernatant was obtained by filtration of the bacterial culture after a 60 min incubation with the test peptide at 100 µm concentration. The results reported in Fig. 9 show that a similar amount of β-galactosidase is released by temporins A and B, a non-significant enzyme quantity is released by temporin H, and no β-galactosidase activity is detectable after incubating bacteria with temporin D.

Figure 9.

Release of β-galactosidase from E. coli D22 cells in the presence of temporins. Cells were incubated in 10 mm sodium phosphate buffer, pH 7.4, for 60 min at 30 °C with 100 µm peptides. Then, cell culture was filtered and 2 mm Gal-ONp was added. The hydrolysis of the β-galactosidase substrate was recorded at 420 nm after 30 min incubation at 37 °C. The bacterial culture without peptide was used as control.

Thus, to investigate the capacity of temporins D and H to cause a weak inner membrane damage, the β-galactosidase activity of the bacterial culture was measured. The results reported in Fig. 10 indicate that only temporin H, although unable to lyse bacteria at that concentration, increased the permeability of the E. coli inner membrane.

Figure 10.

Effect of temporins D and H on the influx of Gal-ONp into E. coli D22. The cells were incubated in 10 mm sodium phosphate buffer, pH 7.4, for 60 min at 30 °C with 100 µm peptides. Gal-ONp (2 mm) was then added to the bacterial culture and its hydrolysis was recorded at 420 nm after 30 min incubation at 37 °C. The bacterial culture without peptide was used as control.

In order to check the cytotoxic effect of temporins, human red cell lysis was studied. The hemolytic activity was found to be highly dependent on the type of assay used. In agar medium, temporins A and B showed an LC value higher than 120 µm[10], whereas temporin D and H were completely inactive. When the capacity to lyse human erythrocytes was tested in liquid medium (0.9% NaCl), the greatest hemolytic activity was obtained with temporin D, which causes more than 80% hemolysis at a concentration above 20 µm(Fig. 11). A similar effect was exhibited by temporins A and B but only at higher peptide concentrations. The differences in activity observed between agar and liquid assays may be related to the differences in solubility and diffusion of the molecules in the two media.

Figure 11.

Percentage hemolysis of human erythrocytes as a function of temporin concentration. Erythrocytes were incubated in 0.9% NaCl for 30 min at 30 °C. The absorbance in the supernatant was recorded at 415 nm. Complete lysis was measured by suspending erythrocytes in distilled water.


Skin secretions of R. temporaria contain antimicrobial peptides belonging to the brevinin-1 and brevinin-2 families [1]. Characteristic of this amphibian species are temporins, recovered from secretions in the range of 14–40 nmol·mg−1 dry weight [10]. Temporin D is the least abundant molecule in the secretion (1 nmol·mg−1), although its identity with the most abundant temporin C is about 84%. On the basis of the retention time in RP-HPLC as well as from the calculated mean residue hydrophobicity, temporin H is the least hydrophobic peptide, but the calculated value for the hydrophobic moment is very similar to that of magainin 2.

Although temporins contain only 13 residues, with the exception of temporin H which is 10 residues long, CD studies indicate that these molecules adopt an α-helical structure in a hydrophobic environment. Moreover, the helical-wheel projections [31] of these peptides show an amphiphilic nature (Fig. 1).

Lipid vesicle permeabilization studies demonstrate that all the peptides are able to lyse artificial membranes, although at different concentrations. The lytic activity of temporins is not affected by the membrane composition, differing from magainins [32] and other antimicrobial peptides [33], which present a lower affinity towards zwitterionic (PtdCho) phospholipid vesicles than towards acidic ones. The lack of selectivity of temporins may be related to the low number of positive charges. In this case, binding to the membrane is mostly due to hydrophobic interactions; such behaviour suggests the occurrence of a barrel-stave mechanism for temporins [34].

The antibacterial assays suggest that both the amphiphilic α-helical conformation and the net positive charge of the peptide are determinant for the bacterial membrane-perturbing ability. In fact, temporin D, which has a hydrophobic moment similar to that of the other temporins but a net charge of +1, is not able to change the permeability properties of the bacterial membrane. The assay performed using LPS-defective strains of E. coli confirms that the low activity of temporins A and B on Gram-negative bacteria is due to their net charge. Once the peptide has reached the bacterial inner membrane, targeted through electrostatic interactions, the hydrophobic interactions play a crucial role in bacterial lysis. Thus, the length of the helix could play an important role in determining the extent of the membrane lesion, as demonstrated by temporin H, which appears to be unable to kill bacteria. In effect, temporin H alters the outer membrane of E. coli D21, inducing a modification in its permeability, as demonstrated by the results obtained following exposure of bacteria to sublethal concentration of rifampicin (Fig. 8). Moreover, the fact that treatment of E. coli D22 with temporin H does not cause the release of β-galactosidase (Fig. 9), although the enzyme substrate Gal-ONp is efficiently hydrolysed inside the bacterial cell (Fig. 10), demonstrates that the inner membrane permeability is also affected.

In conclusion, temporins A, B and H change the permeability properties of the bacterial membranes. According to the results illustrated in Figs 8–10, temporin H renders both the outer and inner membrane permeable to hydrophobic substances of low molecular mass, such as rifampicin and the substrate Gal-ONp, whereas temporins A and B allow the leakage from the cells of larger size molecules, such as β-galactosidase. Temporin D, although devoid of antibacterial activity, has a defined role in the animal defence, being specifically active against eukaryotic cells. Its cytotoxicity against erythrocytes confirms that the hydrophobicity and the relatively low positive charge correlate well with the capacity to perturb the eukaryotic membrane.

The study of the effect of temporins on bacterial cells and artificial vesicles allows important conclusions to be drawn about the minimal structural requirements for lytic efficiency and specificity. This aspect is crucial in economical terms both for the animal, which is challenged with the problem of rapidly synthesizing its own innate immune system against invading microorganisms, and for the opening of new perspectives in the industrial production of new antibiotics.



This work was supported by grants from the Istituto Pasteur-Fondazione Cenci Bolognetti, Consiglio Nazionale delle Ricerche (Target project on Biotechnology, no. 99.00197.PF31 and 99.00289.PF49), and Ministero dell’Università e della Ricerca Scientifica e Tecnologica.


  1. Note: this paper is dedicated to the memory of Professor Vittorio Erspamer, pioneer in the studies of biologically active peptides from amphibian skin, who passed away on 26 October 1999.