Towards the rational design of antimicrobial proteins

Single point mutations can switch on bactericidal and agglutinating activities on the RNase A superfamily lineage


  • David Pulido,

    1. Department of Biochemistry and Molecular Biology, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Spain
    Search for more papers by this author
  • Mohammed Moussaoui,

    1. Department of Biochemistry and Molecular Biology, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Spain
    Search for more papers by this author
  • M. Victòria Nogués,

    1. Department of Biochemistry and Molecular Biology, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Spain
    Search for more papers by this author
  • Marc Torrent,

    Corresponding author
    1. Department of Biochemistry and Molecular Biology, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Spain
    2. Medical Research Council Laboratory of Molecular Biology, Francis Crick Avenue, Cambridge, UK
    • Correspondence

      M. Torrent, Department of Biochemistry and Molecular Biology, Biosciences Faculty, Universitat Autònoma de Barcelona, 08193 Cerdanyola del Vallès, Spain

      Fax: +34 93 581 1707

      Tel: +34 93 581 4147


      E. Boix, Department of Biochemistry and Molecular Biology, Biosciences Faculty, Universitat Autònoma de Barcelona, 08193 Cerdanyola del Vallès, Spain

      Fax: +34 93 581 1264

      Tel: +34 93 581 4147


    Search for more papers by this author
  • Ester Boix

    Corresponding author
    1. Department of Biochemistry and Molecular Biology, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Spain
    • Correspondence

      M. Torrent, Department of Biochemistry and Molecular Biology, Biosciences Faculty, Universitat Autònoma de Barcelona, 08193 Cerdanyola del Vallès, Spain

      Fax: +34 93 581 1707

      Tel: +34 93 581 4147


      E. Boix, Department of Biochemistry and Molecular Biology, Biosciences Faculty, Universitat Autònoma de Barcelona, 08193 Cerdanyola del Vallès, Spain

      Fax: +34 93 581 1264

      Tel: +34 93 581 4147


    Search for more papers by this author


The ribonuclease (RNase) A superfamily lineage includes distant members with antimicrobial properties, suggesting a common ancestral host-defense role. In an effort to identify the minimal requirements for the eosinophil cationic protein (ECP or RNase 3) antimicrobial properties we applied site-directed mutagenesis on its closest family homolog, the eosinophil-derived neurotoxin (EDN or RNase 2). Both eosinophil secretion proteins are involved in human immune defense, and are reported as being among the most rapidly evolving coding sequences in primates. Previous studies in our laboratory defined two regions at the N–terminus involved in the protein antimicrobial action, encompassing residues 8–16 and 34–36. Here, we demonstrate that switching two single residues is enough to provide EDN with ECP antipathogen properties. That is, the EDN double-mutant Q34R/R35W displays enhanced bactericidal activity, particularly towards Gram-negative bacteria, and a significant increase in its affinity towards the bacterial outer membrane lipopolysaccharides. Moreover, we confirmed the direct contribution of residue W35 in lipopolysaccharide binding, membrane interaction and permeabilization processes. Furthermore, additional T13 to I substitution provides EDN with an exposed hydrophobic patch required for protein self-aggregation and triggers bacterial agglutination, thereby increasing the final antimicrobial activity by up to 20–fold. Our results highlight how single selected mutations can reshape the entire protein function. This study provides an example of how structure-guided protein engineering can successfully reproduce an evolution selection process towards the emergence of new physiological roles.


8–aminonaphthalene-1,3,6-trisulfonic acid disodium salt


BODIPY-TR cadaverine




dioleoyl phosphatidylcholine


dioleoyl phosphatidylglycerol


p-xylenebispyridinium bromide


eosinophil cationic protein


eosinophil-derived neurotoxin




minimal inhibitory concentration




Ribonuclease (RNase) A was the first ribonuclease discovered and is the eponym of the vertebrate RNase A superfamily [1, 2], which groups together all the vertebrate RNases homologous to RNase A. In humans, eight functional members have been ascribed to the RNase superfamily and are known as ‘canonical RNases’ [3]. All share common features such as the catalytic triad, composed of two histidines and one lysine, the ability to hydrolyze polymeric RNA substrates, and a specific tertiary structure stabilized by a unique disulfide bond pattern [4]. Notwithstanding, a variety of biological functions have been characterized that are independent of the RNase activity. Specifically, several studies have reported significant antimicrobial activity in some family members, thereby suggesting that RNases might have an ancestral host-defense function [5-8].

The eosinophil ribonucleases, eosinophil-derived neurotoxin (EDN or RNase 2) and eosinophil cationic protein (ECP or RNase 3) are two main secretory components in the secondary granules of eosinophils that contribute to its defense functions. Eosinophils belong to the innate immune system and protect vertebrates from multicellular parasites and certain infections [9]. When activated, eosinophils secrete their contents by degranulation at the infection focus. Mainly, these granules contain cationic proteins and peptides that can directly kill invading pathogens [10]. Eosinophils can also form extracellular traps, as described recently, which trap microorganisms and facilitate their elimination by immune cells, like macrophages [11, 12]. Since around 50 million years ago, EDN and ECP genes have diverged after gene duplication and have rapidly accumulated nonsilent mutations under positive selection pressure, although retaining their folding, cysteine pattern and the catalytic triad essential for ribonuclease activity [13-17]. The antimicrobial and catalytic properties of both proteins have been extensively studied and characterized [10, 18, 19]. EDN was isolated at the beginning of the 1980s as an eosinophil protein of 18.4 kDa able to reproduce the Gordon phenomenon when injected intrathecally into rabbits [20]. Further investigations revealed an intimate relation of EDN with the host immune system by showing that EDN was active against rhinovirus, adenovirus and, most notably, against the respiratory syncytial virus in a ribonuclease-activity-dependent manner [21-24]. However, EDN displays nonsignificant antibacterial [25, 26] and very low antiparasitical activity even at high concentrations [27]. In addition, EDN also acts as a modulator of the host immune system leading to chemotaxis in immature dendritic cells both in vitro and in vivo [28-31]. Therefore, EDN has been classified as an alarmin, a protein capable of activating and modulating the host immune system [31, 32].

By contrast, ECP was first isolated at the beginning of the 1970s, from the secondary granules of human eosinophils as a highly cationic arginine-rich protein. Human ECP is secreted in response to either infection or inflammation processes [33, 34] and displays a highly antimicrobial activity against bacteria [25, 35] and many parasites, such as helminths and protozoa [[19, 27, 36-40]]. In fact, the antimicrobial mechanism of ECP has been extensively studied and, in contrast to EDN, its antimicrobial properties are not dependent on the ribonuclease activity [26].

In spite of all the studies carried out previously, little is known about the molecular determinants that conferred such distinct properties on two proteins that share 65% sequence identity. Despite ECP being more cationic than EDN (pI 10.3 versus 8.6), cationic content by itself is not sufficient to explain the differences observed, because other antimicrobial RNases display pI values close to EDN although retaining antimicrobial activities similar to ECP.

In this study, we investigated the minimal molecular determinants that confer on ECP its unique antimicrobial properties that distinguish it from EDN. Here, we show that by mutating only two residues on EDN, Q34 to R and R35 to W, we can provide EDN with an antimicrobial activity similar to that of ECP. Even more interestingly, we also show that a third mutation, T13 to I, can give to EDN agglutinating properties intrinsically associated with protein self-aggregation in ECP. In summary, in this study we show that a single point mutation on ribonucleases can dramatically enhance the antimicrobial properties and even generate new activities, suggesting that the structure and sequence of RNases may have evolved to embody antimicrobial activity.


Design of EDN mutants

With the aim of studying the molecular determinants of antimicrobial activity in the EDN/ECP lineage, we designed and constructed two EDN mutant proteins using rational mutagenesis. Previous results in our laboratory showed that two regions in the N–terminus of ECP are responsible for the antimicrobial properties: one segment encompassing residues 34–45, responsible for bacterial leakage and membrane permeation; and another segment located in residues 8–16, responsible for protein self-aggregation and bacterial agglutination [18, 41, 42]. Among the former region, it is described that residues R34 and W35 are important for ECP membrane disruptive activity [43], whereas in the latter, region I13 is required for bacteria self-aggregation [44, 45].

To understand whether these residues might have a distinct role in the origin of antimicrobial activity in the ECP lineage, two EDN mutants were obtained: EDN–Q34R/R35W and EDN–T13I/Q34R/R35W. The first incorporates a Trp residue in EDN to generate a RWR membrane-permeabilizing tag similar to that observed in ECP (Fig. 1). The second was designed to include a hydrophobic residue and generate an exposed hydrophobic patch at the surface of the protein which can drive protein self-aggregation and trigger bacterial agglutination (Fig. 1). The parental proteins and mutants were examined using a wide battery of assays to determine the individual contribution of these mutations and to infer the molecular determinants of antimicrobial activity.

Figure 1.

Analysis of the chemical properties of ECP/EDN in antimicrobial-related regions. A molecular surface representation shows the amino acid chemical properties of ECP (left, PDB ID 1QMT [66]) and EDN (right, PDB ID 1GQV [67]). The highly exposed RWR in the antimicrobial region (residues 24–45) is visible for ECP in this representation. In EDN a QRR region is observed instead, highlighting the absence of a hydrophobic residue in this region. In addition, a clear hydrophobic patch in the aggregation region (residues 8–16) is present in ECP, although absent in EDN. Residues were colored according to their chemical properties, cationic residues in blue, anionic residues in red, noncharged polar residues in green, cysteine residues in brown and hydrophobic residues in yellow.

Antimicrobial and cytotoxic activities

ECP, EDN and the two mutants were tested using a minimal inhibitory concentration (MIC) assay for antimicrobial activity against three representative Gram-negative species (Escherichia coli, Acinetobacter baumanii and Pseudomonas sp.) and three representative Gram-positive species (Staphylococcus aureus, Micrococcus luteus and Enterococcus faecium) (Table 1). As previously described [42], ECP displayed high antimicrobial activity in a submicromolar range against both Gram-positive and Gram-negative bacteria, whereas EDN showed only minor or undetectable antimicrobial activity for the bacterial species tested. By contrast, both EDN mutants Q34R/R35W and T13I/Q34R/R35W displayed antimicrobial activity on all strains and enhanced the antimicrobial activity up to 20–fold with respect to EDN. More interestingly, mutation of EDN Thr13 to Ile enhanced the antimicrobial activity even more, thus reflecting the contribution of the aggregation region to the protein antimicrobial mechanism.

Table 1. Bactericidal (MIC100) and hemolytic activity (HC50) of ECP, EDN and EDN mutants.
E. coli Pseudomonas sp. A. baumanii S. aureus M. luteus E. faecium
  1. a

    Hemolytic activity was assayed on sheep red blood cells.

ECP0.> 25
EDN> 10109> 10> 109> 25
EDN-Q34R/R35W10.> 25
EDN-T13I/Q34R/R35W0.> 25

Further investigations on the bactericidal properties of the EDN mutants were compared with ECP by assaying the membrane depolarization activity against two bacterial model strains E. coli and S. aureus (Table 2). As previously described [46], ECP is able to interact with the Gram-negative and Gram-positive bacterial envelope, and can perturb the cell cytoplasmic membrane, producing half of the total membrane depolarization at concentrations of 0.04 μm on E. coli and 0.05 μm on S. aureus. By contrast, its homolog protein EDN did not show any depolarization activity on E. coli on S. aureus at the concentrations assayed. However, Q34R/R35W and T13I/Q34R/R35W EDN mutants displayed high depolarization activity against both Gram-positive and Gram-negative strains, showing an effect similar to ECP.

Table 2. Depolarization activity on E. coli and S. aureus cells determined by DiSC3(5) assay for ECP, EDN and EDN mutants. n.d., not detected at the assayed concentration range (0.1–10 μm).
 Depolarization (μm)
E. coli S. aureus
  1. a

    Maximum fluorescence value reached at the final incubation time.

ECP0.040 ± 0.0011459 ± 3080.048 ± 0.0093628 ± 289
EDN> 10n.d.> 10n.d.
EDN-Q34R/R35W0.027 ± 0.0041218 ± 1210.033 ± 0.0022285 ± 289
EDN-T13I/Q34R/R35W0.025 ± 0.0041381 ± 1550.039 ± 0.0052262 ± 173

Cytotoxicity against eukaryotic cells was tested using the hemolytic assay by incubating erythrocytes with serially diluted concentrations of ECP, EDN and the corresponding mutants (Table 1). Notably, none of the four proteins tested displayed any hemolytic activity below 25 μm, suggesting that the mutations specifically increased the antimicrobial activity of the protein. Our results, which are also supported by cytotoxic activity measurements on different human cell lines [43, 47], differ from those reported by Young et al. in 1986 [48]. In that study, the authors observed high hemolytic activity for ECP preparations from eosinophils. These discrepancies might be attributed to other proteins or peptides from eosinophils retained as impurities in their ECP purification method.

Lipopolysaccharide binding affinity and cell agglutinating activity

Lipopolysaccharide (LPS) binding affinity was monitored using a cadaverine (BODIPY-TR cadaverine; BC) fluorescent probe, which measures the competitive displacement of BC by the proteins tested. The LPS-binding affinities of ECP, EDN and the corresponding mutants were tested by comparing their ED50 values, defined as the concentration for which half BC displacement occurs (Table 3). EDN binding to LPS was reduced 10–fold with respect to ECP. However, EDN mutants Q34R/R35W and T13I/Q34R/R35W showed an enhanced LPS-binding affinity between four and five times higher than their parental protein EDN. It is interesting to note that after protein addition at 5 μm, BC was completely displaced by ECP, whereas EDN reduced the occupancy by only 50% at the same concentration. More significantly, both EDN mutants decreased the occupancy by ~ 70%, approaching the ECP effect (Table 3). EDN mutants binding to LPS were also confirmed using LPS micelles by recording the Trp fluorescence emission (Table 4), suggesting that only ECP and EDN mutants were able to interact with the Gram-negative outer membrane molecule. In addition, the registered wavelength shifts for ECP and the two mutants pointed to a direct contribution of W35 residue in the LPS interaction.

Table 3. Minimal agglutination activity (MAC) and LPS binding (ED50) of ECP, EDN and mutants.
 MAC (μm)LPS binding
E. coli Pseudomonas sp. A. baumanii S. aureus M. luteus E. faecium ED50m)%maxa
  1. n.d., not detected. a 100% refers to a total displacement whereas 0% is for no displacement of the dye, indicating no binding.

ECP0.60 ± 0.050.60 ± 0.051.50 ± 0.05n.d.n.d.n.d.0.86 ± 0.13100
EDNn.d.n.d.n.d.n.d.n.d.n.d.7.49 ± 1.0648
EDN-Q34R/R35Wn.d.n.d.n.d.n.d.n.d.n.d.1.76 ± 0.3374
EDN-T13I/Q34R/R35W4.50 ± 0.054.50 ± 0.054.50 ± 0.05n.d.n.d.n.d.1.82 ± 0.4165
Table 4. Tryptophan fluorescence in the presence of lipid vesicles, LPS and SDS micelles for ECP, EDN and mutants.
 Trp fluorescencea
λm (nm)λs (nm)λm (nm)λs (nm)λm (nm)λs (nm)λm (nm)λs (nm)
  1. a

    λm represents the wavelength of the maximum and λs the shift in the wavelength compared with the buffer.

3 : 2 DOPC/DOPG3427341334153404

Previous results from our group described that LPS binding is required to agglutinate bacteria [49]. In order to quantify the agglutinating activity, we determined the minimal concentration of protein that was able to agglutinate bacterial cells, defined as the minimal agglutination concentration. ECP, EDN and their derived mutants were evaluated after incubation with all strains tested for antimicrobial activity (Table 3). Neither EDN nor EDN–Q34R/R35W was able to agglutinate bacteria at the concentrations tested. However, EDN carrying the I13A mutation was the only mutant able to trigger bacteria agglutination in Gram-negative cells. The absence of agglutination in Gram-positive bacteria highlights that the presence of a LPS outer layer is required for agglutination. These results suggest that agglutination can only be triggered when an exposed hydrophobic patch is present.

To further investigate the agglutinating activity of the EDN mutants, we analyzed bacterial cultures using SEM to evaluate cell-surface morphology and aggregate size and density (Fig. 2). As expected, after incubation of E. coli with EDN, cells retain their baton-shaped morphology and do not display significant damage. In turn, both EDN mutants promote severe envelope damage but only the T13I mutation shows bacteria agglutination, registered as dense aggregates with an average size of 10 μm length.

Figure 2.

SEM micrographs for E. coli cultures incubated in the absence and presence of ECP, EDN and EDN mutants Q34R/R35W and T13I/Q34R/R35W). Two magnifications (upper and lower panels) are shown for each condition to visualize the extent of bacteria aggregates and cell morphology.

Interaction with model membranes

To further characterize the antimicrobial mode of action in the designed mutants, we tested their ability to interact with lipid vesicles and disrupt their structure.

To monitor the protein–lipid interaction, we monitored the intrinsic tryptophan fluorescence signal and the λmax shift displacement induced by changes on the residue microenvironment. We recorded the protein tryptophan fluorescence spectrum before and after incubation with charged (dioleoyl phosphatidylglycerol; DOPG), neutral (dioleoyl phosphatidylcholine; DOPC) and a mixture of charged and neutral (DOPC/DOPG) liposomes, and also in the presence of SDS micelles (Table 4). When the spectrum was recorded in the presence of liposomes containing negatively charged phospholipids or SDS micelles, a significant blue shift was observed for both ECP and EDN mutants, but no appreciable shift was detected for EDN. Previous characterization of both ECP Trp structural properties revealed that whereas W10 is mostly embedded inside the protein core, W35 is hyperexposed to the solvent and can partially insert into lipid bilayers [50]. For its part, EDN shares with ECP the buried W10 but lacks W35, showing an additional Trp at the N–terminus, W7, which is mainly exposed to the solvent. Surface-accessible areas calculated using areaimol (CCP4i) were: W7 (132 Å2) W10 (9 Å2) for EDN and W10 (30 Å2) and W35 (252 Å2) for ECP. Therefore, the experimental data indicate that EDN Trp residues are not involved in membrane binding, whereas insertion of a Trp at position 35 is enhancing the protein lipid interaction [51].

Finally, to characterize the lipid bilayer destabilization properties of EDN and its mutants we assayed the disruption of 8–aminonaphthalene-1,3,6-trisulfonic acid disodium salt/p-xylenebispyridinium bromide (ANTS/DPX) containing large unilamellar vesicles, where an increase in fluorescence can be recorded as the vesicle leakage proceeds (Table 5). EDN did not promote leakage in the presence of either charged or neutral lipid vesicles at the concentrations tested. However, EDN mutants were able to induce high leakage activity on DOPG-containing vesicles, with IC50 values similar to ECP. It is also worth noting that no leakage was recorded for DOPC micelles for either ECP- or EDN-derived mutants at 10 μm, showing again that electrostatic interactions drive the protein–lipid interaction process.

Table 5. Liposome leakage of ANTS/DPX LUVs by ECP, EDN and EDN mutants.
 Liposome leakage (ED50 in μm)
ECP0.046 ± 0.0090.14 ± 0.08> 10
EDN> 10> 10> 10
EDN-Q34R/R35W0.025 ± 0.0060.080 ± 0.004> 10
EDN-T13I/Q34R/R35W0.029 ± 0.0080.030 ± 0.007> 10


The eosinophil ribonucleases, EDN and ECP, are two secretory proteins from secondary granules of eosinophils. Dating back 50 million years, after divergence of the Old World monkeys from the New World monkeys, ECP and EDN precursors originated following a gene duplication event, and have been accumulating nonsilent mutations until now [13, 16]. Nonsilent mutations of EDN/ECP accumulated faster than all the other known functional coding sequences among primates, probably because of a positive selection pressure [17, 52]. Many mutations, especially those increasing the cationic nature of ECP, have been linked to its increased antimicrobial action, but we cannot ascertain which substitutions on the ancestral precursor were key in ECP developing antimicrobial activity.

Using peptide synthesis approaches, we have proven that the entire protein is not required to display high antimicrobial activity [41, 42]. In this context, we found that a protein N–terminal fragment conserves both the antimicrobial and agglutinating properties of ECP. These findings allowed us to locate specific properties to particular regions of the protein [18]. We identified a key antimicrobial region (residues 24–45) that is essential for membrane leakage, depolarization and LPS binding. Recent work on ECP-binding domains also highlighted the involvement of the 34–38 stretch for both heterosaccharide and lipid binding [53, 54]. A second region, close to the N–terminus of the protein (residues 8–16) was found to be essential for bacteria agglutination [45].

When these regions are examined for sequence conservation (Fig. 3), we observe that the antimicrobial region in both proteins is fairly well conserved, excluding a highly surface-exposed group of amino acids (34–36). Therefore, this region may have been selected by evolution to explore a potential antimicrobial domain in ribonucleases. If we compare the chemical properties of the residues in the aforementioned region (Fig. 3), we can see a distinct pattern for EDN and ECP. Whereas EDN has clear charge and hydrophobic segregation, ECP has scrambled both residue types along this region. This observation is in agreement with the interfacial activity model [55], in which a limited rather than an accurate amphipathicity in proteins and peptides would confer on them antimicrobial capacity. When we mutate residues Q34 to R and R35 to W, we generate a similar charge and hydrophobic distribution in EDN as observed in ECP, giving this mutant an activity similar to that of ECP (Table 1).

Figure 3.

Conservational analysis of EDN/ECP. A molecular representation of ECP (left, PDB ID 1QMT [66]) and EDN (right, PDB ID 1GQV [67]) shows the amino acid conservation score during evolution (upper) and the chemical properties (lower) of the antimicrobial region. The figure shows a backbone of conserved residues both for ECP and EDN. The amino acids analyzed in this study are among residues displaying high sequence variability. Interestingly, the latter residues can change the vectorial chemical properties of the protein. By breaking down the highly segregated properties of EDN into a more scrambled structure, we could efficiently generate antimicrobial activity in EDN. Conservation scores were calculated using Consurf and colored according to the conservation score: very high (purple), high (pink), low (dark blue) or very low (light blue) degree of conservation. For chemical representation, cysteines (brown), cationic (blue), anionic (red), non-charged polar (green) and hydrophobic (yellow) amino acids were colored.

If this is the case, we can compare different antimicrobial ribonucleases and check for analogous distributions [18]. In the RNase A superfamily, we can see then that antimicrobial RNases, such as RNase 3 and 7, show an imperfect distribution of hydrophobic and charged residues (Fig. S1). However, nonantimicrobial RNases, such as RNase 1 and 2, do not show this pattern and/or accumulate negatively charged residues that mask positive charges.

The fact that only one or two mutations are needed to change this pattern shows that ribonucleases have a suitable scaffold for becoming antimicrobial proteins, suggesting that this may not be a collateral product of evolution but a true original function of ribonucleases. Indeed, close inspection of the conserved sequence pattern in eosinophil RNases (Fig. S2) highlights unambiguous divergence at the selected mutated residues.

Similar conclusions are reached when we analyze the agglutinating properties of the EDN/ECP lineage. A closer look at the chemical properties of these regions shows that ECP has a more hydrophobic region than EDN (Fig. 3), which may explain why only ECP has agglutinating activity. In fact, in our results, only one single-point mutation, T13 to I, was enough to develop agglutinating properties in EDN (Table 3). We propose that this is due to the generation of a potent exposed hydrophobic patch near the N–terminus of the protein. The conservation analysis of these regions shows that, as with the antimicrobial segment, the amino acid residues are overall well conserved (Figs 3 and S2). Nonetheless, the T13 by I substitution, which increases the hydrophobic character of the region considerably, is enough to provide an agglutinating activity to EDN, although it cannot fully reproduce the action of ECP (Table 3). This might be because EDN has a glutamic residue in position 12 that may hinder the protein–protein interaction required for bacteria agglutination.

Indeed, comparing this region among antimicrobial ribonucleases of the RNase superfamily, we identify a positive correlation between agglutinating activity and the hydrophobicity of the aforementioned patch (Fig. S3). We again observe a structural pattern in ribonucleases displaying antimicrobial-related properties, reinforcing the idea that ribonucleases are a suitable molecular scaffold for the development of antimicrobial proteins.

Ribonucleases are not an exception among antimicrobial proteins. Other proteins with activities unrelated to antimicrobials have also been reported and a similar behavior could be observed, in which targeted point mutations triggered or abolished the antimicrobial activity [56]. However, this study shows that evolution does not need to shape an entire protein or even a full domain to develop antimicrobials, but single substitutions in suitable exposed regions, or even selected post-translational modifications can modulate the antimicrobial properties [57]. It should not be surprising, therefore, that a growing number of proteins not even closely related to host-defense mechanisms might display antimicrobial activity [[58, 59]], as found for some apolipoproteins [60] or the Alzheimer beta peptide [61].

It is therefore of general interest to pursue the investigation into the RNase A family lineage, which may assist us in unraveling the structural and sequence requirements for a protein to become antimicrobial.

Experimental procedures

Materials and strains

DOPC and DOPG were from Avanti Polar Lipids (Alabaster, AL, USA). ANTS, DPX and BC were purchased from Invitrogen (Carlsbad, CA, USA). Lipoteichoic acids from S. aureus and LPS from E. coli serotype 0111:B4 were purchased from Sigma-Aldrich (St. Louis, MO, USA). PD-10 desalting columns with Sephadex G-25 were from GE Healthcare (Waukesha, WI, USA). Strains used were E. coli (BL21; Novagen, Madison, WI, USA), S. aureus (ATCC 502A), A. baumannii (ATCC 15308), Pseudomonas sp. (ATCC 15915), M. luteus (ATCC 7468) and En. faecium (ATCC 19434).

Expression and purification of recombinant proteins

Wild-type ECP and EDN were obtained from human synthetic genes. All mutations were incorporated using the Quick-change mutagenesis kit from Invitrogen and sequenced before expression. Protein expression in the E. coli BL21DE3 strain, folding of the protein from inclusion bodies and purification were carried out as previously described [62, 63]. Protein identity was checked by MALDI–TOF MS. CD spectra were recorded to confirm that the introduced mutations did not perturb the native protein overall 3D structure.


Antimicrobial activity was calculated as the MIC, defined as the lowest protein concentration that completely inhibits microbial growth. The MIC of each protein was determined from two independent experiments performed in triplicate for each concentration. Proteins were dissolved in 10 mm sodium phosphate buffer, pH 7.5, and serially diluted from 10 to 0.2 μm. Bacteria were incubated at 37 °C overnight in Luria–Bertani broth and diluted to give ~ 5 × 105 CFU·mL−1. In each assay, protein solutions were added to each bacteria dilution and incubated for 4 h, and samples were plated onto Petri dishes and incubated at 37 °C overnight.

Minimal agglutination activity

Bacterial cells were grown at 37 °C to D600 = 0.2, centrifuged at 5000 g for 2 min and resuspended in 10 mm Tris/HCl buffer, 0.1 m NaCl, pH 7.5. An aliquot of 100 μL of the bacterial suspension was treated with increasing protein concentrations (0.01–10 μm) and incubated at room temperature for 1 h. The aggregation behavior was observed by visual inspection and the agglutinating activity is expressed as the minimum agglutinating concentration of the sample tested, as previously described [64].

Bacteria cytoplasmic membrane depolarization assay

Membrane depolarization was followed using the method described previously [64]. Briefly, bacteria strains were grown at 37 °C to D600 = 0.2, centrifuged at 5000 g for 7 min, washed with 5 mm Hepes at pH 7.2 containing 20 mm glucose, and resuspended in 5 mm Hepes–KOH, 20 mm glucose, and 100 mm KCl at pH 7.2 to D600 = 0.05 and 200 μL was transferred to a microtiter plate. DiSC3(5) was added to a final concentration of 0.4 μm and changes in the fluorescence were continuously recorded after addition of protein (5 μm) in a Victor3 plate reader (Perkin–Elmer, Waltham, MA, USA). The time required to achieve total membrane depolarization and half membrane depolarization (IC50) was estimated from nonlinear regression analysis as previously described [46]. Lysis of the cells with the detergent Triton X–100 gives maximum membrane depolarization.


Cultures of E. coli and S. aureus (1 mL) were grown at 37 °C to the mid-exponential phase (D600 = ~ 0.4) and incubated with proteins (5 μm) in NaCl/Pi at room temperature. Sample aliquots (500 μL) were taken after up to 4 h of incubation and prepared for SEM analysis as previously described [64]. The micrographs were viewed at an accelerating voltage of 15 kV on a Hitachi S-570 SEM, and a secondary electron image of the cells for topography contrast was collected at several magnifications.

Fluorescent probe displacement assay for LPS and lipid A binding to ECP

LPS binding was assessed using the fluorescent probe BC as previously described [65]. BC binds strongly to native LPS specifically recognizing the lipid A portion. When a protein that interacts with LPS is added, BC is displaced from the complex and its fluorescence is increased, decreasing its occupancy factor. LPS-binding assays were carried out in a 5 mm Hepes buffer at pH 7.5. The displacement assay was performed in a microtiter plate containing a stirred mixture of both LPS (10 μg·mL−1) and BC (10 μm). Proteins were serially diluted from 10 to 0.1 μm. Fluorescence measurements were performed on a Victor3 plate reader. The concentration required to achieve half lipopolysaccharide binding (ED50) was estimated from nonlinear regression analysis as previously described [46].

Hemolytic activity

Fresh human red blood cells were washed three times with NaCl/Pi (35 mm phosphate buffer, 0.15 m NaCl, pH 7.4) by centrifugation for 5 min at 3000 g and resuspended in NaCl/Pi at 2 × 107 cells·mL−1. Red blood cells were incubated with proteins at 37 °C for 4 h and centrifuged at 13 000 g for 5 min. The supernatant was separated from the pellet and its absorbance measured at 570 nm. The 100% hemolysis was defined as the absorbance obtained by sonicating red blood cells for 10 s. HC50 was calculated by fitting the data to a sigmoidal function.

Liposome preparation

Large unilamellar vesicles of ~ 100 nm diameter were prepared from a chloroform solution of DOPC, DOPG or a DOPC/DOPG mixture (3 : 2 molar ratio). After vacuum-drying, the lipid film was suspended in 10 mm Tris/HCl, 0.1 m NaCl, pH 7.4 buffer to give a 1 mm solution, then frozen and thawed several times prior to extrusion through polycarbonate membranes as previously described [50].

Fluorescence measurements

Tryptophan fluorescence emission spectra were recorded using a 280 nm excitation wavelength. Slits were set at 2 nm for excitation and 5–10 nm for emission. Emission spectra were recorded from 300 to 400 nm at a scan rate of 60 nm·min−1 in a 10 × 10 mm cuvette, with stirring immediately after sample mixing. Protein spectra at 0.5 μm in 10 mm Hepes buffer, pH 7.4, were obtained at 25 °C in the absence or presence of 200 μm liposome suspension, 1 mm SDS micelles or 100 μm LPS micelles. Fluorescence measurements were performed on a Cary Eclipse spectrofluorimeter (Agilent Technologies, Bath, UK). Spectra in the presence of liposomes were corrected for light scattering by subtracting the corresponding large unilamellar vesicle background. For each condition three spectra were averaged. The maximum of the fluorescence spectra was calculated by fitting the data to a log-normal distribution function as previously detailed [50].

ANTS/DPX liposome leakage assay

The ANTS/DPX liposome leakage fluorescence assay was performed as previously described [50]. Briefly, a unique population of large unilamellar vesicles was prepared to encapsulate a solution containing 12.5 mm ANTS, 45 mm DPX, 20 mm NaCl and 10 mm Tris/HCl, pH 7.5. The ANTS/DPX liposome stock suspension was diluted to 30 μm and incubated at 25 °C with proteins, serially diluted from 10 to 0.1 μm in a microtiter plate. Fluorescence measurements were performed on a Victor3 plate reader. IC50 values were calculated by fitting the data to a dose-response curve.


Transmission and scanning electron microscopy were performed at the Servei de Microscopia of the Universitat Autònoma de Barcelona (UAB). Spectrofluorescence assays were performed at the Laboratori d'Anàlisi i Fotodocumentació, UAB. We thank Vivian A. Salazar for her contribution in the preparation of the graphical material. EDN plasmid for recombinant protein expression was kindly provided by Richard J. Youle (NINDS, NIH, Bethesda, MD, USA). The work was supported by the Ministerio de Educación y Cultura (grant number BFU2009-09371) and Ministerio de Economía y Competitividad (BFU2012-38965), co-financed by FEDER funds and by the Generalitat de Catalunya (2009 SGR 795). DP is a recipient of a FPU predoctoral fellowship (Ministerio de Educación y Cultura) and MT is a recipient of a Beatriu de Pinós fellowship (Generalitat de Catalunya).