Research article

A new RNase sheds light on the RNase/angiogenin subfamily from zebrafish

Elio Pizzo, Antonello Merlino, Mimmo Turano, Irene Russo Krauss, Francesca Coscia, Anna Zanfardino, Mario Varcamonti, Adriana Furia, Concetta Giancola, Lelio Mazzarella, Filomena Sica, Giuseppe D'Alessio


Recently, extracellular RNases of the RNase A superfamily, with the characteristic CKxxNTF sequence signature, have been identified in fish. This has led to the recognition that these RNases are present in the whole vertebrate subphylum. In fact, they comprise the only enzyme family unique to vertebrates. Four RNases from zebrafish (Danio rerio) have been previously reported and have a very low RNase activity; some of these are endowed, like human angiogenin, with powerful angiogenic and bactericidal activities. In the present paper, we report the three-dimensional structure, the thermodynamic behaviour and the biological properties of a novel zebrafish RNase, ZF-RNase-5. The investigation of its structural and functional properties, extended to all other subfamily members, provides an inclusive description of the whole zebrafish RNase subfamily.

  • angiogenin
  • bactericidal activity
  • RNase
  • vertebrate
  • zebrafish (ZF) (Danio rerio)


Only recently, when a large number of fish DNA and protein sequences has become available, has it been possible to recognize in fish proteins the unmistakable signature of an RNase sequence (CKxxNTF). Thus it has become clear that RNases from the ‘RNase A superfamily’ are present not only in tetrapods, but also in fish, and hence in all vertebrates. This has led to the proposal to rename the superfamily as the ‘vertebrate RNase superfamily’ [1], a proposal in line with the relevant finding that a single family of enzymes has been recognized as present exclusively in the subphylum of vertebrates, that of RNases [2].

Several fish RNases have been identified and studied: four recombinant RNases from ZF (zebrafish) (Danio rerio) [35] and two from salmon (Salmo salar) [6]. All fish RNases studied so far have a weak RNA-degrading activity, and most of them are effectively angiogenic. As in the case of hANG (human angiogenin), their angiogenic activity is strictly dependent on the integrity of the enzyme catalytic activity [5,6]. On the other hand, all fish RNases studied so far have been found to be endowed with a bactericidal activity [3,6]. Surprisingly, however, the latter activity, when investigated with salmon RNases, has been found to be maintained when the RNases are either catalytically inactivated or fully denatured [6].

In the last decade, hANG has been assigned various key biological roles, besides that of an angiogenic effector, such as regulation of proliferation of endothelial and cancer cells, stimulation of rRNA transcription, stimulation of neurite outgrowth and pathfinding of motor neurons [7], and control of stress-induced translational arrest [8,9].

ZF-RNases have been independently studied in different laboratories. Three ZF-RNases (ZF-RNase-1, -2 and -3) have been reported by Pizzo et al. [5]; two of them, plus a third different enzyme, have been studied by Cho and Zhang [3]. Kazakou et al. [4], after an extensive survey of ZF DNA sequences, which indicated widespread polymorphism among the known ZF-RNase homologues, produced the X-ray structure of two of the RNases (RNase ZF-1a and -3e). An assessment of the nomenclatures of the available ZF RNases is presented in Table 1, which includes a novel ZF-RNase-5 member, encoded by a gene sequence described by Kazakou et al. [4] as due to polymorphism (see below). The nomenclature of ZF-RNases as first proposed by Pizzo et al. [5], and its extension, will be used throughout the present paper.

View this table:
Table 1 Nomenclatures of ZF-RNases

The expression during development and in adult animals of some ZF-RNases has been investigated by different laboratories, with diverse results [3,10]. This approach is of special interest, given the possibility that ZF offers a model experimental system for studies of angiogenesis development, an approach that has already been successfully explored [1113].

A recent study has revealed that ZF-RNase-1 and -2 are able to induce phosphorylation of extracellular-signal-regulated kinase 1/2–mitogen-activated protein kinase [14]. Furthermore, they undergo nuclear translocation, as hANG does, and accumulate in the nucleolus, where they stimulate rRNA transcription.

Given the interest in the diverse bioactivities of ZF-RNases and the expediency of ZF as a model experimental system, particularly for angiogenesis studies, we deemed it of interest to prepare and characterize a novel ZF-RNase, ZF-RNase-5. Based on its amino acid sequence, the gene encoding this RNase has been previously proposed as a polymorphic variant of the gene encoding ZF-RNase-2 [4]. We investigated the three-dimensional structure of the protein, its thermodynamic behaviour and its biological profile, and concluded that ZF-RNase-5 is a protein and effector distinct from ZF-RNase-2 and all other ZF-RNases. As the investigation of ZF-RNase-5 properties has been extended to those of the other ZF-RNases, an inclusive description has emerged of the whole ZF-RNase subfamily.


Cloning, expression and purification of recombinant ZF-RNases

ZF-RNases -1, -2 and -3 were produced as described in [5]. A plasmid vector containing the gene sequence encoding ZF-RNase-4 was generously provided by Dr Jianzhi Zhang, Department of Ecology and Evolutionary Biology, University of Michigan. To amplify the mature portion of the gene sequence, the following primers were used: 5′-TCCATATGCAGTCTTATAATGACTTCAAACGC-3′ (forward) and 5′-CCCAAGCTTTTAAGAATTGTTGGAACGTCCATA-3′ (reverse); underlining indicates restriction sites added to perform cloning. The gene sequence encoding ZF-RNase-5 was obtained by PCR using as a template a cDNA clone (IMAGp998L1415615Q) purchased from imaGenes. To amplify the gene sequence, the following primers were used: 5′-GAAATTCCATATGAAGGTTCCACCAGACGTA-3′ (forward) and 5′-CCCAAGCTTTATTTAGCCTGACCTGTTTAC-3′ (reverse); underlining indicates restriction sites added to enable cloning. The signal peptides of ZF-RNase-4 and ZF-RNase-5 were predicted using the SignalP 3.0 server (

The PCRs were performed under the following conditions: an initial denaturation step at 94 °C for 5 min, followed by 35 cycles of 2 min at 94 °C, 2 min at 62 °C and 1 min at 73 °C. PCR products were isolated by electrophoresis on a low-melting 1% agarose gel, and purified using the High Pure PCR Product Purification Kit (Roche Applied Science, Monza, Italy). Purified DNA, treated with NdeI and HindIII restriction enzymes, was inserted into a pET22b(+) expression vector (Novagen) and purified using a MiniPrep extraction kit (Qiagen). All cloned and purified DNAs were certified through sequencing (MWG Biotech) before processing.

The recombinant expression plasmids, with the exception of pET22b(+)/ZF-RNase-4, were used to transform competent Escherichia coli strain BL21(DE3) (Invitrogen). The expression plasmid encoding ZF-RNase-4 was used to transform competent E. coli strain C43(DE3) (Invitrogen). Cells were grown at 37 °C to a D600 of 1 and then induced with 0.1 M isopropyl β-D-thiogalactopyranoside and grown overnight.

The purification of the five recombinant ZF-RNases was carried out as described previously [5]. Briefly, inclusion bodies were solubilized in 7 M GdnHCl (guanidinium chloride), 100 mM Tris acetate, pH 8.4, containing 1 mM glutathione. After 10 min of nitrogen flushing, each preparation was left for 2 h at room temperature (20 °C). Renaturation was obtained through an initial dilution of 1:20, drop by drop, in 100 mM Tris acetate, pH 8.4, containing 0.5 M L-arginine and 1 mM oxidized glutathione. After 24 h at room temperature, each preparation was dialysed against 50 mM Tris/HCl, pH 7.4, and loaded on to an SP-Sepharose column (GE Healthcare) equilibrated in the same buffer. Elution was carried out using a gradient from 0 to 1 M NaCl in the same buffer. The fractions containing ZF-RNases were revealed through SDS/PAGE (15% gel) followed by zymograms and pooled. To remove residual contaminants, each fraction pool was loaded on to a reverse-phase C-4 column (Phenomenex) equilibrated in 100% solution A [5% (v/v) acetonitrile and 0.1% trifluoroacetic acid]. The column was eluted with a gradient in which the concentration of solution B [composed of 90% (v/v) acetonitrile containing 0.1% trifluoroacetic acid] was raised to 100% in 1 h. For each preparation a single major protein component was eluted, which by SDS/PAGE was found to contain a single protein.

Assays of bactericidal activity

Bacterial strains of E. coli strain DH5α, Pseudomonas fluorescens (A.T.C.C. 13525), Staphylococcus aureus (A.T.C.C. 6538P) and Bacillus subtilis (PY79) were used in the antibacterial tests, performed as described previously [15]. The bacteria were grown overnight, diluted 1:1000 in 20 mM sodium phosphate buffer, pH 7.0, and incubated with 3 μM (final concentration) ZF-RNase at a density of 4000 cfu (colony-forming units)/ml. After 6 h at 37 °C, serial dilutions of each protein/bacteria mixture were prepared and plated, and the cfu remaining after each treatment were determined. Negative controls were carried out with proteins from E. coli strain BL21(DE3) transformed with an empty pET22b(+) vector. For each experiment, carried out in duplicate, triplicate assays were performed. S.D.s were 4–10%, or as detailed for each experiment. LD50 values (the concentrations required for a 50% bactericidal effect) were determined by bactericidal assays at increasing concentrations up to 2 μM for each of the five RNases.

RT (reverse transcription)–PCR experiments

Temporal expression patterns of ZF-RNases were determined by RT–PCR using total RNAs extracted from embryos at different developmental stages with the RNeasy Protect mini kit (Qiagen). Total RNA (1 μg) from each time point was reverse-transcribed to cDNA with the Qiagen OneStep RT–PCR kit, following the manufacturer's instructions. cDNAs were amplified using the following specific primers: for ZF-RNase-1, 5′-TTTATTCATAACGCTGCTTTTCA-3′ (forward) and 5′-CCATGTTGCCCTGTGGAC-3′ (reverse); for ZF-RNase-2, 5′-TCACAACAGTGCTGTTCAATACA-3′ (forward) and 5′-TTCGTGTCGACCACCCGAT-3′ (reverse); for ZF-RNase-3, 5′-ACTTACGGTCAACCAGCAGAA-3′ (forward) and 5′-ACAAGTCTCTGTTATCAGTTTGTCG -3′ (reverse); for ZF-RNase-4, 5′-CCTGGGTTTAATTACCAGATTTC-3′ (forward) and 5′-CAATTATGTTTTCTATTTTGGTGCAG-3′ (reverse); for ZF-RNase-5, 5′-AAAAGACAACACTTCAGTCAAAAGG-3′ and 5′-AAACAGGTTGTTCACCGGAG-3′ (reverse); and for ZF-odc1 (ornithine decarboxylase 1; a housekeeping gene), 5′-TTGCAATCAAATCTTGAACAAA-3′ (forward) and 5′-GGAGGTGCTTCTTCAGGACA-3′ (reverse). All pairs of gene-specific primers were engineered to anneal to two different exons, so that cDNA amplifications could be distinguished by product size and from genomic contaminant amplifications. PCRs were carried out using 1 μl of cDNA as a template, 10 pmol of each primer and 1.25 units of Taq DNA polymerase (Fermentas) in a total volume of 25 μl, under the following conditions: one cycle at 94 °C for 5 min, followed by 35 cycles (40 cycles for ZF-RNase-2 and -4) of 20 s at 94 °C, 30 s at 60 °C and 40 s at 72 °C, with a final cycle at 72 °C for 7 min.

Angiogenesis assays

HUVECs (human umbilical vein endothelial cells), cultured in EBM-2 (endothelial basal medium) containing the EGM-2 (endothelial growth medium) Bullet kit (Cambrex), were seeded in Matrigel™-coated 48-well plates (Becton Dickinson) at a density of 4×104 cells per well in 150 μl of EBM-2. ZF-RNase-4, ZF-RNAse-5 and recombinant hANG were added to the cells at different concentrations and incubated at 37 °C for 4 h. Cells were fixed with phosphate-buffered glutaraldehyde (0.2%) and paraformaldehyde (1%), and photographed.

Assays of RNase activity

Zymogram assays of RNase activity were carried out using SDS/PAGE electropherograms as described previously [16]. Ribonucleolytic assays in vitro were carried as described in [17] with the fluorigenic substrate 6-carboxy-fluorescein-dArUdAdA-6-carboxy-tetramethylrhodamine (Integrated DNA Technologies). The assay mixtures contained 0.1 M imidazole/HCl, pH 6.0, 0.1 M NaCl, 20–60 nM substrate and suitable enzyme aliquots. All preparations were checked for constant specific activity with the in vitro assay indicated above before using them for any biological or physicochemical test.

Chemical and physical denaturation of ZF-RNases

Reduced and alkylated ZF-RNases were obtained by treating ZF-RNases (1 mg/ml) for 2 h at 37 °C with a 10-fold molar excess of dithiothreitol in 200 mM Mes/NaOH, pH 6.0. Carboxymethylation of the exposed sulfhydryl groups was carried out by adding a 20-fold molar excess (with respect to total −SH concentration) of iodoacetamide. After 60 min at room temperature in the dark, the protein was freed of excess reagents and by-products by gel filtration through a PD10 column (GE Healthcare Bio-Sciences) equilibrated in 0.1 M ammonium acetate at pH 5.0, and was lyophilized. MS analyses revealed an increase in molecular mass consistent with the effects of the carboxymethylation reaction.

Thermal denaturation of ZF-RNases was achieved by treating the proteins for 5 min at 85 °C, followed by rapid cooling to 0 °C.

CD spectroscopy

CD measurements were carried out on a Jasco J-715 spectropolarimeter equipped with a Peltier-type temperature control system (model no. PTC-348WI). Molar ellipticity per mean residue, [Θ] in degrees·cm2·dmol−1, was calculated from the equation [Θ]=[Θ]obs·(mrw/10lC), where [Θ]obs is the ellipticity measured in degrees, mrw is the mean residue molecular mass (117 Da), l is the optical pathlength of the cell in centimetres and C is the protein concentration in moles per litre. A 0.1-cm pathlength cell and a protein concentration of 0.2 mg/ml in 10 mM sodium phosphate buffer, pH 7.2, were used. CD spectra were recorded at 25 °C in a 0.1-cm quartz cell, averaged over three scans from 260 to 190 nm with a time constant of 16 s, a 2-nm band width and a scan rate of 20 nm·min−1. CD spectra were baseline-corrected by subtracting the buffer spectrum. The instrument was calibrated with an aqueous solution of D-10-camphorsulfonic acid at 290 nm [18]. Thermal unfolding was monitored in the temperature scan mode at 222 nm from 25 °C to 85 °C with a scan rate of 1 °C·min−1.

The fraction of denatured protein (ΔfD) was calculated as ΔfD=(Θ–Θmin)/(Θmax–Θmin); Θ is the ellipticity at a given temperature, and Θmax and Θmin are the maximum and minimum values of ellipticity corresponding to the denatured and native states of the proteins respectively.

The urea-induced transition curves were obtained by recording the ellipticity at 222 nm as a function of denaturant concentration. Measurements were performed after overnight incubation of samples at 4 °C. The protein concentration was 0.2 mg/ml. The enthalpy changes were calculated by fitting CD melting curves from the van't Hoff equation using the Origin 7.5 program (OriginLab).

Fluorescence spectroscopy

Intrinsic protein fluorescence was recorded using a Jasco FP-750 spectrofluorimeter equipped with a circulating water bath. The excitation wavelength was set at 280 nm, and the emission was measured between 300 and 450 nm. The spectra were recorded at room temperature with a 1-cm cell and a 5-nm emission slit width, and were corrected for background signal.

The urea-induced transition curves were obtained by recording fluorescence intensity at the wavelength maximum as a function of denaturant concentration. As for CD measurements, fluorescence measurements were performed after overnight incubation of samples at 4 °C at a protein concentration 0.2 mg/ml.

DSC (differential scanning calorimetry)

DSC measurements were carried out on a third-generation Setaram Micro-DSC instrument. A scanning rate of 0.5 °C·min−1 was chosen for all experiments. Protein solutions (2 mg/ml) were prepared in 10 mM sodium phosphate buffer, pH 7.2. Raw data were converted into an apparent molar heat capacity, taking into account the instrument calibration curve and the buffer–buffer scanning curve, and by dividing each data point by the scan rate and the protein molar concentration in the sample cell. Reheating runs were repeated to determine the calorimetric reversibility of the thermal-denaturation process. The excess molar heat capacity function 〈ΔC0p〉 was obtained after baseline subtraction, assuming as a reference the heat capacity of the native state [19]. The denaturation enthalpies, ΔH0cal, were obtained by integrating the area under the heat capacity against temperature curves. Tm is the melting temperature and corresponds to the maximum of each DSC peak. The entropy changes ΔS0cal were determined by integrating the curve obtained from dividing the heat capacity curve by the absolute temperature. The denaturation enthalpies, entropies and Gibbs energies at 298 K were calculated according to the classical Kirchhoff equations.

Crystallization and diffraction procedures

Crystals of ZF-RNase-1 were obtained by the hanging-drop vapour diffusion method. Drops were prepared at 20 °C by mixing equal volumes of protein solution (15 mg/ml) and reservoir solution containing 1.8–2 M ammonium sulfate and 0.1 M sodium acetate, pH 4.2–4.8 (ZF1_pH4.5). To analyse the structure of ZF-RNase-1 at higher pH, the crystals, grown at pH 4.5, were transferred to harvesting solutions (2.3 M ammonium sulfate) buffered at pH values in the range 4.5–7.3. A stepwise procedure in which the pH was adjusted in steps of 0.2 was necessary to avoid deterioration of crystal quality. Crystals were stored under the final pH condition (ZF1_pH7.3) for 1 week with daily changes of the mother liquor.

The best crystals of ZF-RNase-5 were obtained using the hanging-drop vapour diffusion method by mixing equal volumes of protein (15 mg/ml) and precipitant solution containing 32% (w/v) mPEG [methoxypoly(ethylene glycol)] 2000, 0.1 M sodium acetate, pH 4.5, and 0.2 M ammonium sulfate.

Diffraction data of ZF1_pH4.5 were collected by using synchrotron light at ELETTRA, Trieste, Italy. In the other two cases, data were collected on a Saturn944 CCD (charge-coupled device) detector with a Cu-Kα radiation from a Rigaku Micromax 007 HF generator. In all cases, after the addition of 30% (w/v) glycerol to the harvesting solution, crystals were flash-frozen at 100 K in supercooled nitrogen gas produced by an Oxford Cryosystem instrument and maintained at 100 K during data collection. All data were indexed, processed and scaled using the HKL2000 package [20].

Structure determination

The structure of ZF1_pH4.5 was solved by molecular replacement, using the Phaser program [21] and the model of the homologous bovine angiogenin (PDB code 1AGI) [22], as the structure of the protein determined at pH 8.0 (ZF1_pH8) [4], was not yet available. The Phaser program was also employed to solve the structure of ZF-RNase-5 using ZF1_pH4.5 as the starting model.

Refinement was carried out using CNS (Crystallography & NMR System) software (version 1.2) [23]. Each refinement step was followed by manual interventions using the program O [24] to correct minor errors in the positions of some side chains and to identify solvent sites. The final models of ZF1_pH4.5, ZF1_pH7.3 and ZF-RNase-5 have an R-factor (Rfree) of 0.162 (0.192), 0.152 (0.205) and 0.178 (0.210) respectively, and have been deposited in the PDB (codes 3LJD, 3LN8 and 3LJE respectively.). Statistics and parameters of the refinements are given in Table 2.

View this table:
Table 2 Data collection and refinement statistics

Values in parentheses correspond to the highest resolution shells.

All structures were validated with PROCHECK [25] and WHATCHECK [26]. Cartoons were generated using PyMOL (DeLano Scientific;

Homology modelling

The structures of ZF-RNase-2 and ZF-RNase-4 were modelled with the SWISS-MODEL server [27], using as a reference the crystal structures of Zf-RNase-5 and ZF-RNase-3 respectively. The templates were chosen on the basis of the highest score and the lowest E-value found with a BLAST search against the PDB database. Positional sequence identity between ZF-RNase-2 and ZF-RNase-5 was 73%, whereas that between ZFRNase-4 and ZF-RNase-3 was 33%. The models produced by the server were then manually modified to build missing residues on the N-terminal α-helical region and energy-minimized in vacuo by means of the GROMOS96 force-field, following a procedure previously reported [28,29].

The stereochemical quality of the final models showed no residues positioned in the disallowed regions of the Ramachandran plot. The Z-score values of the combined statistical potential energy of the ZF-RNase-2 and ZF-RNase-4 models were −6.04 and −5.12 respectively. These values are in the range of scores typically found in proteins of similar sequence length and are analogous to the values of the templates, which are equal to −6.22 and −6.30 respectively.

The Z-score indicates overall model quality and measures the deviation of the total energy of the structure with respect to an energy distribution derived from random conformations [30,31].


The primary structure of ZF-RNase-5

The expression of ZF-RNase-5, and its purification as a recombinant protein, are described in the Experimental section. The protein sequence is illustrated in Figure 1, in comparison with the sequence of all other ZF-RNases, hANG and RNase A. ZF-RNase-5 contains in its primary structure the classical RNase signature (CKxxNTF), and the histidine and lysine residues essential for catalysis are located at positions corresponding to those present in the other active RNases from the vertebrate superfamily. There are 34 amino acid substitutions in the sequence of ZF-RNase-5 with respect to that of ZF-RNase-2, ten of which are conservative. This gives a 73% identity, higher than the identity values between ZF-RNase-5 and the other ZF-RNases, which are in the range 30–60%.

Figure 1 Amino acid sequences of ZF-RNases

Residues shared by all five ZF-RNases and by human angiogenin are shown in black with a grey background, whereas the residues shared also by RNase A are shown in white with a black background.

The three-dimensional structure of ZF-RNase-5

The three-dimensional structure of ZF-RNase-1 (ZF1_pH8) and that of ZF-RNase-3, obtained at basic pH, have been described in [4]. We focused our attention on the structure of the newly identified ZF-RNase-5. Furthermore, we inspected new crystal forms of ZF-RNase-1, and analysed ZF-RNase-2 and -4 by homology modelling. The structural models are shown in Figure 2, whereas the superimposition of their Cα atoms is reported in Supplementary Figure S1 at

Figure 2 Ribbon representation of the structures of ZF-RNases

The crystal model of ZF-RNase-5 (residues 1–121) was refined to an R-factor of 17.8% (Rfree 21.0%) at 1.8 Å resolution (1 Å=0.1 nm). A summary of the refinement statistics is presented in Table 2. In comparison with the previously described ZF-RNases (-1 and -3) [4] and hANG (PDB code 1ANG), the RMSD (root mean square deviation) of the corresponding Cα atoms are 0.70 Å, 0.91 Å and 1.28 Å respectively.

We found that, among the ZF proteins, the most relevant differences are located in helix II. In ZF-RNase-5, this helix, which encompasses residues 24–35, has a first turn in a 310 conformation and terminates with a turn in a π-helix conformation with main-chain hydrogen bonds Met31–Ile36 and Ser32–Lys35 and Lys35 in the Lα conformation. In the corresponding region (residues 23–36) of ZF-RNase-1 [4], the network of hydrogen bonds (Ile30–Ile37 and Gly31–Lys36) is spatially conserved. Lys36 is in the Lα conformation, and the intervening extra residues Pro32 and Asn33 bulge out of the helix, forming a type I β-bend (Supplementary Figure S2 at Thus the topology of this fragment is substantially unchanged with respect to ZF-RNase-5, although the insertion of the two residues formally breaks the helix at the level of Ile30 (Figure 2 and Supplementary Figure S3 at In the case of ZF-RNase-3 (residues 23–35) [4], the first turn of the helix is highly distorted.

ZF-RNase-5 is the first structure of a ZF-RNase that presents a sulfate ion in P1; its interactions with the surrounding catalytic residues (His17, His117 and Lys45) and structured water molecules are similar to those found in RNase A (see e.g. [32] and other angiogenins [33,34]) (Figure 3A). The similarity of the active-site architecture to that of RNase A extends to the subsite P2, where Lys12 and Arg15 are the analogues of Lys7 and Arg10 in the bovine pancreatic enzyme. However, in ZF-RNase-5, the B1 subsite delimited by Val47, Thr49, and Tyr118 is obstructed by the part of the C-terminus that is visible in the electron-density map (residues 119–122). In particular, the side chain of Glu120 makes two hydrogen bonds with Thr49 and partly mimics the binding interactions of the pyrimidine base. In the present case, the orientation of the glutamic acid side chain is enforced by the involvement of this residue in a pseudo type II′ β-bend stabilized by a C10 hydrogen bond between the side chain of the preceding Asp119 and Gly121 (Figure 3B).

Figure 3 Structural features of ZF-RNase-5

(A) Active-site region of ZF-RNase-5. The structure of ZF-RNase-5 (green) is superimposed on that of RNase A (pink) in complex with the ATAA tetranucleotide (purple). (B) Superimposition of the C-terminal region of ZF-RNase-5 (green) and RNase A (pink). The pseudo type II′ β-bend and the C10 hydrogen bond between Asp119 and Gly121 in ZF-RNase-5 are shown.

In order to investigate whether the different behaviour in the sulfate binding between ZF-RNase-5 and ZF-RNase-1 and -3 was not merely a result of differences in the environmental parameters used in the crystallization trials, we extended our analysis to crystals of ZF-RNase-1 grown at acidic pH and in the presence of an elevated concentration of SO42− ions. This produced a new crystal form that contains two molecules in the independent unit (chains A and B). The crystals obtained at pH 4.5 were also equilibrated at pH 7.3, and the structures at the two pH values were refined independently (ZF1_pH4.5 and ZF1_pH7.3). Note that the recombinant protein lacks the first three residues in the sequence [4]. To simplify the comparison with ZF1_pH8, the numbering of the latter has been maintained. The electron-density map is very well-defined for most of the protein residues. A summary of the refinement statistics is presented in Table 2. The Cα atomic positions for the two molecules in the asymmetric unit do not differ significantly and their RMSD is 0.42 Å at pH 4.5 and 0.41 Å at pH 7.3.

The main structural differences between ZF1_pH4.5 and ZF1_pH7.3 are limited to solvent molecules and to the conformation of a few side chains. Inspection of the electron-density maps of ZF1_pH4.5 reveals that chain A binds five sulfate ions and chain B binds three sulfates and one acetate ion. Details of the interactions of these ions with the protein are reported in Supplementary Table S1 at The binding site of a sulfate moiety (an acetate group in molecule B) and that of a second sulfate group are well conserved in both chains. The remaining non-equivalent sites are involved in packing contacts. The map of the anionic binding sites is strictly conserved in the structure at neutral pH, the only exception being the acetate moiety, which is replaced by an additional sulfate anion. Thus, in contrast with numerous crystallographic reports on RNase A [32, 35] and other members of the family [3638], and angiogenins [33,34], and despite the relatively high concentration in the crystallization mixture, none of the sulfate ions is located in the active site, as the anion binds preferentially to other regions of ZF-RNase-1. The putative binding subsite B1 of the pyrimidine base is partially obstructed by the side chain of Glu122 located in the C-terminal segment of the protein. The position of this residue, fixed by a hydrogen bond to Thr52, is common to all of the angiogenins that have a glutamic acid residue in the equivalent position and, in particular, to ZF1_pH8 [4]. Interestingly, the C-terminal segment, to which the glutamic acid residue belongs, is also the region where the largest differences are observed between the structure at basic pH and those at lower pH. In the latter, this region is better defined, probably due to packing interactions involving residues 125–127. Overall, the C-terminus causes an obstruction of B1 that is even greater than that produced in ZF1_pH8

Physicochemical properties of ZF-RNase-5 and the other ZF-RNases

The conformational stability of recombinant ZF-RNase-5 was investigated by CD and DSC and compared with those of the other ZF-RNases, hANG and non-angiogenic RNase A, the RNase superfamily prototype.

When the thermal unfolding of the RNases was monitored by CD spectroscopy at 222 nm (Figure 4), melting curves with sigmoid profiles were obtained. The Tm of ZF-RNase-5 was found to be close to those of ZF-RNase-2, ZN-RNase-4, hANG and RNase A, whereas ZF-RNase-1 and -3 exhibited melting temperatures approx. 8–9 °C lower (Table 3). The denaturation enthalpy changes, ΔH0v·H, calculated from the CD melting curves with the van't Hoff equation (see the Experimental section) are shown in Table 3. The equation describes two-state N↔D transitions, where N and D represent the native and the denatured state respectively.

Figure 4 Far-UV CD melting profiles recorded at 222 nm

A ΔfD value represents the fraction of denatured protein (see the Experimental section). Symbols are as indicated in brackets: ZF-RNase-1 (triangles), ZF-RNase-2 (squares), ZF-RNase-3 (stars), ZF-RNase-4 (vertical bars), ZF-RNase-5 (circles), RNase A (broken line), hANG (solid line).

View this table:
Table 3 Physicochemical parameters of ZF-RNases, hANG and RNase A

C1/2 values were the urea concentrations at half-completion of the transition. The error in Tm is ±1 °C and ±0.5 °C for CD and DSC measurements respectively. The error in ΔH0cal and ΔS0cal is <5%. The error in ΔS0v·H·l is <10% and in ΔG0298 <15%.

By DSC measurements, good agreement was established between the calculated Tms and those found in CD experiments, as well as between the denaturation enthalpies ΔH0cal directly measured from the DSC curves, and those measured with the van't Hoff equation (Table 3). These data confirmed that the thermal denaturation process is a two-state transition process for all of the proteins. The enthalpy value for ZF-RNase-5 was found to be close to those of ZF-RNase-1 and -3 and lower than the value obtained for hANG (approx. 50 kJ·mol−1). The enthalpy values for ZF-RNase-2 and -4 were instead close to the ΔH0cal of hANG and lower than the value found for RNase A.

In addition, as the denaturation of the RNases, followed by CD or DSC, was found to be a reversible process at thermodynamic equilibrium, the entropy and Gibbs energy could be calculated. In Supplementary Figure S4 at, CD spectra and CD melting profiles illustrate the reversibility of the unfolding processes. Reversibility of unfolding for RNase A [39] and human angiogenin [40] has been described previously.

The entropy change values showed the same trend as the enthalpy change values (Table 3): the ΔS0cal value of ZF-RNase-5 was determined to be close to that of ZF-RNase-1 and -3 and lower than the value obtained for hANG, whereas ΔS0cal values of ZF-RNase-2 and -4 were determined to be close to that of hANG. An inspection of Gibbs energy values, ΔG0298 (listed in Table 3), showed that the thermodynamic stability of ZF-RNase-5 and of all ZF-RNases falls in the range 21–29 kJ·mol−1, a stability close to that of hANG and far from that of RNase A.

The conformational stability of ZF-RNases against the denaturing action of urea was also investigated using two different spectroscopic methodologies: CD, which reflects conformational changes of the secondary structure, and steady-state fluorescence, to analyse conformational changes of the tertiary structure.

Very similar values of 1/2, the urea concentration at half-completion of transition, were obtained by CD and fluorescence measurements for each protein, as shown in Table 3. The results of these experiments indicated that a simultaneous collapse of both secondary and tertiary structures occurs following denaturation. Furthermore, it confirms that the urea-induced unfolding of all of the proteins is a two-state denaturation process, involving only native and denatured states. The value of 1/2 of ZF-RNase-5 was comparable with that of ZF-RNase-4 and hANG, whereas the 1/2 of ZF-RNase-2 was found to be the highest among the ZF-RNases. The values of 1/2 of ZF-RNase-1 and -3 were found to be the lowest among the ZF-RNases investigated.

In conclusion, the stability against urea follows the same trend as thermal stability and confirms that ZF-RNase-1 and -3 have less-stable native structures. They also indicate that ZF-RNase-5 and -2 have significantly different structural stabilities, thus suggesting that they are distinct proteins.

The ribonucleolytic, angiogenic and bactericidal activities of ZF-RNase-5

ZF-RNase-5, like the other fish RNases studied so far [36], is characterized by a low ribonucleolytic activity. This is not surprising given its ability to act as an angiogenin (see below) and the typical low RNase activity of angiogenins. For measuring the activity, a continuous, very sensitive, assay was chosen [17], previously proposed [41] as particularly suitable for measuring the low activity values of angiogenins and for a comparison of the activities of several RNases/angiogenins.

When the activity of ZF-RNase-5 was compared with those of the other ZF-RNases, hANG and RNase A by performing assays in parallel, ZF-RNase-5 was found to be much less active than RNase A, the superfamily prototype, and appromimately as moderately active as hANG and the other ZF-RNases (Table 4). It should be noted that ZF-RNase-5 was found to be two orders of magnitude more active than ZF-RNase-2 [kcat/Km=(1.25 ± 0.09)×104 compared with kcat/Km=(1.91 ± 0.13)×102].

View this table:
Table 4 Kinetic data on the ribonucleolytic activity of ZF-RNases, hANG and bovine pancreatic RNase A, the prototype of the vertebrate RNase superfamily

With a standard two-dimensional assay, ZF-RNase-5 was found to be as angiogenic as hANG, as shown in Figure 5. The results of angiogenic assays on ZF-RNase-4, never assayed previously, are also shown in Figure 5. As we have reported for the previously investigated ZF-RNases [5,6], we found also that ZF-RNase-5 and -4 lost their angiogenic activity when their RNase activity was obliterated (results not shown).

Figure 5 Angiogenic activity of ZF-RNase-4 and -5

Assays were carried out with primary HUVECs. (A) Negative control obtained with non-supplemented EBM-2 medium. (B) Human angiogenin (200 ng/ml). (C and D) Recombinant ZF-RNase-4 tested at 200 and 400 ng/ml. (E and F) Recombinant ZF-RNase-5 tested at 200 and 400 ng/ml.

ZF-RNase-5 was found to possess a strong bactericidal activity on Gram-negative bacteria (E. coli and Pseudomonas aeruginosa were tested) and no activity on Gram-positive S. aureus and B. subtilis (Figure 6). Identical results were obtained with ZF-RNase-4, with LD50 values in the 0.75–1.0 μM range for both RNases. In contrast, ZF-RNases -1 and -3 were found to possess a low to weak, but significant, activity also on Gram-positive bacteria, and ZF-RNase-2 was found to be very active on these bacteria. The latter finding is of interest because ZF-RNase-5, proposed to be a variant of ZF-RNase-2, has no activity on Gram-positive bacteria.

Figure 6 Bactericidal activity of ZF-RNases (3 μM final concentration) after incubation of bacterial cells for 6 h

Cell survival was determined, as a percentage of controls, with untreated bacteria or bacteria treated with 3 μM BSA. Results shown are means, n=3; S.E.M was lower than 5%. The bacteria used were: (A) E. coli; (B) P. fluorescens; (C) B. subtilis; (D) S. aureus.

It may not be surprising that the bactericidal activity of ZF-RNase-5, like that of the other ZF-RNases, is conserved when the ribonucleolytic activity of the RNases is suppressed by alkylation of the catalytically essential histidine residues (see the Experimental section). However, it is surprising that their bactericidal activity is also conserved when the structure of the RNases is unfolded, either by heating or by reduction of the protein disulfides and alkylation of the freed sulfhydryls [6,42] (results not shown).

ZF-RNase-5 in development

Given that ZF is a most convenient experimental model, especially for embryogenesis studies, and particularly for angiogenesis, and given the availability of the ZF angiogenin genes and proteins described in the present paper, we investigated the temporal expression profile of ZF-RNase-5 during embryonic/larval development. We found that the RNase was found to be expressed at all stages of development. As conflicting results have been reported for the other previously investigated ZF-RNases [3,10], we decided to assay or re-assay all the known ZF-RNases in parallel.

As illustrated in Figure 7, we confirmed the lack of expression during development of ZF-RNase-2 and -4, but found a strong expression of ZF-RNase-1 and -3, especially for the gene encoding ZF-RNase-3 at 48 h after fertilization. In particular, the results in Figure 7 show that ZF-RNase-2 has no role in embryogenesis, whereas ZF-RNase-5, purportedly a variant of ZF-RNase-2, is present during the whole period of development.

Figure 7 Expression patterns of ZF-RNases during ZF embryo development

RT–PCR assays were performed. Total RNA, extracted from whole-body embryos at the indicated times post-fertilization, was treated with DNase and retrotranscribed into cDNA. No expression was detectable for ZF-RNase-2 and -4. ZF-RNase-1, -3 and -5 were expressed at all stages analysed. The expression of the housekeeping gene (ZF-odc1) was used as a positive control.


To date, four members of the vertebrate RNase superfamily have been studied, and partly characterized, in the ZF subfamily: ZF-RNase-1, -2, -3 and -4, all isolated as recombinant proteins [35,10,14]. A fifth member, ZF-RNase-5, is proposed in the present paper to be a cell product by itself, rather than a polymorphic variant of ZF-RNase-2 as previously suggested [4]. This proposal is supported by several lines of evidence. (i) The enzymatic activity of ZF-RNase-5 is two orders of magnitude higher than that of ZF-RNase-2. (ii) The bactericidal activity on Gram-positive bacteria is absent in ZF-RNase-5, but quite strong in ZF-RNase-2. (iii) ZF-RNase-2 is not expressed during development, whereas ZF-RNase-5 is strongly expressed in the whole developmental period. (iv) The denaturation enthalpic and entropic values determined for ZF-RNase-2 are close to those of hANG, whereas those determined for ZF-RNase-5 are much lower. Urea denaturation experiments also indicate that ZF-RNase-2 is much more stable than ZF-RNase-5.

ZF-RNase-5 has been crystallized, and its structure, determined by X-ray crystallography, has been directly compared with that of ZF-RNase-2 and -4, deduced through modelling, and the crystallographic models of ZF-RNase-1 and -3. With the due caution required when comparing structural and functional differences between proteins having significant differences in the amino acid sequence, some general trends can be extracted from the structural features of the different members of the subfamily.

Based on their structural homology with RNase A, putative substrate-binding subsites, previously identified in angiogenins, can also be identified for ZF-RNases. The B1 and B2 subsites interact specifically with bases, whereas P0, P1 and P2 subsites interact with phosphate groups. Although the geometry of the P1 subsite is strictly conserved among ZF-RNases, it should be noted that only ZF-RNase-5 binds a sulfate group in this subsite. This result is enforced by the fact that ZF-RNase-5 crystals were grown from solutions with a low ammonium sulfate concentration (0.2 M). This feature is probably related to a cluster of positive charges (Arg9, Lys12 and Arg15) in the proximity of the ZF-RNase-5 active site. In particular, Lys12 and Arg15 correspond to Lys7 and Arg10 of RNase A, two residues that line the P2 subsite and are considered to be important for substrate binding [43].

An important role in determining the low catalytic activity of angiogenins in comparison with RNase A has been ascribed to the obstruction of the B1 subsite that occurs in most angiogenins [22,33,44]. In these molecules, a glutamic acid/glutamine residue on the C-terminal tail protrudes in the B1 subsite and forms two hydrogen bonds with the threonine residue involved in pyrimidine base recognition. This feature is also observed in ZF-RNase-5. In this protein, as for hANG, the glutamic acid residue is involved in a pseudo type II′ β-bend stabilized by a hydrogen bond between the amide nitrogen of Gly121 and the side chain of Asp119. This structural motif, which further restricts the conformational space available to the glutamic acid residue, is not found in ZF-RNase-1, where the aspartic acid residue is replaced by a histidine residue. In this respect, it should be recalled that the corresponding aspartate residue of RNase A (Asp121) forms a sort of ‘catalytic dyad’ with His119 [45]. In the bovine pancreatic enzyme, the hydrogen bond between the side chains of these two residues stabilizes the histidine conformation that is supposed to be active during the transphosphorylation step. In ZF-RNase-5, just as in hANG, no interactions between the corresponding aspartic acid residue and histidine residue are observed and, in the two enzymes, aspartic acid is engaged in interactions with Ser118 and Gly121 respectively. However, it should be noted that a rotation of the aspartic acid side chain from the t to the g conformation would bring the carboxylate group in a position to form the hydrogen bond with the catalytic histidine residue. Moreover, the rotation of the aspartic acid side chain also results in the disruption of the hydrogen bond with the second residue along the chain (Gly121 in ZF-RNase-5). This would also destabilize the position of the C-terminal residue and favour the expulsion of Glu120 from the B1 subsite (Figure 3B). This structural feature, together with the greater ability of ZF-RNase-5 to bind a sulfate anion in the active site, may account for the enhanced enzymatic activity of the latter with respect to ZF-RNase-1.

It should be noted that the most relevant differences between ZF-RNases and hANG are localized at the C-terminal tail. In hANG, this region adopts a well-defined 310-helix conformation, which is anchored to the protein core by hydrophobic interactions involving residues downstream of the obstructive Gln117 (Ile119 and Phe120). In ZF-RNases, the C-terminal segment seems to be much more mobile. Indeed, in ZF1_pH8 it is partially disordered and the part visible in the electron-density map assumes a different conformation when compared with that observed in the structures of ZF1_pH4.5 and ZF1_pH7.3 reported in the present study. In fact, in ZF-RNase-5 the C-terminus appears even more disordered: the last visible residue in the electron-density map is Gly121. This could be reflected in an increased capability of ZF-RNase-5 to adopt an alternative conformation of the C-terminal tail that allows the binding of substrate.

Interestingly, despite ZF-RNase-2 sharing 73% sequence identity with ZF-RNase-5, its catalytic activity is 100-fold lower. This result could be ascribed to a lower number of basic residues in the active-site region of ZF-RNase-2 in comparison with ZF-RNase-5. Moreover, the putative B2 subsite of the former enzyme appears to be less accessible as a result of the presence of Arg107 replacing Gly109 in the latter. Concerning ZF-RNase-4, the most active member of the family of ZF-RNases, its high activity could be due to the concomitant occurrence of an unobstructed B1 subsite, a feature found also in ZF-RNase-3, and of a cluster of positive charges at the N-terminal helix. In the neighbourhood of the catalytic histidine residue (His10), this cluster could help in substrate recognition.

The structural differences noted among the different members of the ZF-RNase subfamily, in particular at the C-terminal region, approximately correlate with their thermodynamic behaviour. The presence of the pseudo type II′ β-bend in ZF-RNase-2 and ZF-RNase-5, but not in ZF-RNase-3 and ZF-RNase-1, may be the source of their slightly different thermal stability. This feature is not present in the modelled structure of ZF-RNase-4, which also has a shorter N-terminal helix. Its stability, comparable with that of ZF-RNase-2 and ZF-RNase-5, may well be due to the compactness of the protein and, in particular, to the shorter length of the loop connecting helix II to the β-strand that is next in the sequence, a feature that the enzyme shares with hANG and RNase A.

As for the bactericidal activity of ZF-RNases, we have completed the inspection of all known ZF-RNases and found that they are all bactericidal, particularly on Gram-negative bacteria. However, the most relevant result is the finding that their bactericidal activity is preserved not only after inactivation of their ribonucleolytic activity, but also after denaturation of the proteins. Clearly, the finding that bactericidal RNases do not use their ribonucleolytic activity for killing bacteria is not new, as it has also been reported for other bactericidal RNases [15,46,47]. Certainly surprising is instead the finding that they can perform a biological function, such as host defence, a mechanism of innate immunity, while devoid of any structure. This question has been previously discussed for salmon RNases [6], which present the same behaviour. Apparently, the proteins can bind to and penetrate the bacterial membrane also when unfolded, possibly for the presence in their amino acid sequences of many positively charged residues, which could form clusters that can readily act on the negatively charged outer leaflet of the bacterial membrane [46].

The finding of RNases in the phylogenetically distant fish proteome, and their ability to exert both angiogenic and bactericidal activity, has led to different proposals: that the earliest function of RNases in evolution was to promote angiogenesis [1], or to act as host-defence effectors [3]. In fact, it cannot be excluded – in a gene-sharing fashion [48] – that ancestral RNases were involved in both angiogenesis and host defence. Certainly surprising is the conclusion that these ancestral host-defence effectors evolved into the ordered RNase structure, although they did not require this structure to exert their activity; on the other hand, the angiogenic activity could be exerted only by exploiting an RNase active site inside a well-ordered RNase structure.

In conclusion, the results of the present study on the structure of ZF-RNases, their thermodynamic stability, and on the relationships between their structure and their RNase activity, as well as the results on their bactericidal and angiogenic activity, and on their expression early in development, provide an enlarged, more comprehensive, picture of the ZF-RNases, by adding many tesserae towards a complete mosaic.


Irene Russo Krauss, Francesca Coscia and Antonello Merlino performed the crystallization and X-ray data-collection experiments, solved and refined the crystal structures and performed the homology modelling calculation. Concetta Giancola performed and analysed the experiments addressed at determining the stability of the proteins. Elio Pizzo performed the preparation of recombinant RNases and the assays of RNase activity and angiogenic activity. Mario Varcamonti and Anna Zanfardino designed and performed the assays of bactericidal activity. Mimmo Turano designed and performed the determination of the expression of RNases in development. Filomena Sica, Lelio Mazzarella, Concetta Giancola, Adriana Furia and Giuseppe D'Alessio designed the research and wrote the paper. All authors discussed the manuscript.


This work was supported by the Ministero Italiano dell'Università e della Ricerca Scientifica (PRIN 2007).


We thank Giosuè Sorrentino and Maurizio Amendola for technical assistance, Elettra Trieste for providing synchrotron radiation facilities, Dr Antimo Di Maro for MS analyses and Gaetano D'Amato for preparing ZF-RNAs.


  • The structural co-ordinates reported will appear in the PDB under accession codes 3LJD, 3LN8 and 3LJE.

Abbreviations: cfu, colony-forming units; DSC, differential scanning calorimetry; EBM, endothelial basal medium; hANG, human angiogenin; HUVEC, human umbilical vein endothelial cell; odc1, ornithine decarboxylase 1; RMSD, root mean square deviation; RT, reverse transcription; Tm, melting temperature; ZF, zebrafish


View Abstract