HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Petrakis, T. G. Articles by Talianidis, I. Search for Related Content PubMed PubMed Citation Articles by Petrakis, T. G. Articles by Talianidis, I. Social Bookmarking                      What's this?

J Biol Chem, Vol. 274, Issue 35, 24726-24730, August 27, 1999

Cloning and Characterization of Mouse Deoxyguanosine Kinase
EVIDENCE FOR A CYTOPLASMIC ISOFORM^*

Thodoris G. Petrakis^§, Eleni Ktistaki, Liya Wang^¶, Staffan Eriksson^¶, and Iannis Talianidis

From the  Institute of Molecular Biology and Biotechnology, Foundation for Research and Technology Hellas, P. O. Box 1527, 711 10 Herakleion Crete, Greece and the ^¶ Department of Veterinary Medical Chemistry, The Biomedical Center, S-751 23 Uppsala, Sweden

	ABSTRACT

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Deoxyguanosine kinase (dGK) is a nuclear gene product that catalyzes the phosphorylation of purine deoxyribonucleosides and their analogues. The human enzyme is located predominantly in the mitochondria, as shown by biochemical fractionation studies and in situ localization of the overexpressed recombinant protein. Here we describe the cloning of mouse dGK cDNA and the identification of a novel amino-terminally truncated isoform that corresponds to about 14% of the total dGK mRNA population in mouse spleen. In situ fluorescence assays suggest that the new isoform cannot translocate into the mitochondria and thus may represent a cytoplasmic enzyme. Expression of mouse dGK mRNA was highly tissue-specific and differed from the tissue distribution observed in humans. Recombinant mouse dGK showed similar specific activity and substrate specificity as compared with the human enzyme. The broad specificity, restricted tissue distribution, and location of mouse dGK in multiple cellular compartments raise new considerations with respect to the role of the individual deoxynucleoside kinases in nucleotide metabolism.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

In mammalian cells, the synthesis of dNTP¹ precursors for nuclear DNA synthesis is performed by two pathways: (a) the de novo pathway, and (b) the salvage pathway. The rate-limiting step in the de novo pathway, which involves the initial synthesis of ribonucleotides from ribose-5-phosphate, amino acids, and CO₂, is the irreversible removal of the 2'-hydroxyl group from ribonucleotide diphosphates, catalyzed by ribonucleotide reductase (1). The alternative pathway for dNTP synthesis involves phosphorylation of deoxynucleosides derived from either endogenous dNTP breakdown or their uptake from the extracellular space. In this pathway, the rate-limiting reaction is the first phosphorylation step catalyzed by cytoplasmic deoxynucleoside kinases (2). Because the dNTP pools produced by both pathways are in rapid equilibrium, the actual contribution of the de novo or salvage pathway for nuclear DNA synthesis depends on several factors such as cell type, cell cycle, or the physiological state of the cell (3, 4).

In contrast to the nucleus, the inner membrane of the mitochondria is impermeable to charged molecules such as deoxyribonucleotides. There is no definitive evidence for the existence of enzymes involved in the de novo pathway inside the mitochondria. Therefore, the main supply of dNTPs for DNA synthesis in this organelle seems to be deoxynucleoside salvage regulated by two recently cloned mitochondrial enzymes: (a) deoxyguanosine kinase (dGK), and (b) thymidine kinase (TK)-2 (5-8). Human dGK can efficiently phosphorylate deoxyguanosine and deoxyadenosine, whereas TK2 phosphorylates deoxyadenosine and deoxycytidine (5-8). This suggests that the combined action of the two enzymes is sufficient for the synthesis of all four dNTPs for mitochondrial DNA replication. Human dGK and dCK have overlapping but not identical substrate specificities toward both natural substrates and therapeutic nucleoside analogues (5, 6, 9-11). For example, certain analogues, such as 2',2'-difluorodeoxyguanosine, are phosphorylated mainly by dGK, whereas others, such as 2-chloro-2'-deoxyadenosine or 9--D-arabinofuranosyl-guanine, are substrates for both dCK and dGK. These characteristics are important for understanding the toxic side effects of some widely used anti-viral drugs and raise a new concept regarding the development of drugs specifically targeted to interfere with mitochondrial DNA replication. An in-depth understanding of the regulation and biochemical properties of this enzyme is therefore of particular interest, not only in its basic biological aspect, but also for further pharmacological exploitation.

Here we report the cloning, expression, and functional characterization of mouse dGK cDNA. We present evidence for the expression of a cytoplasmic isoform of mouse dGK. This activity may play a role in dNTP synthesis for nuclear DNA replication or repair. In addition, it may have important implications in the evaluation of the contribution of mitochondrial or nuclear DNA damage induced by certain drugs in mouse models.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Isolation of Mouse dGK cDNA Clones-- Approximately 1 × 10⁶ plaques of a mouse spleen cDNA library in ggt11 (CLONTECH) were screened with a cDNA probe encompassing the entire coding region of human dGK (5). Hybridization was performed overnight at 65 °C in 6× saline/sodium phosphate/EDTA (1× saline/sodium phosphate/EDTA = 0.15 M NaCl and 10 mM sodium phosphate/1 mM EDTA), 5× Denhardt's solution (1× Denhardt's solution = 0.02% each of bovine serum albumin, polyvinylpyrrolidone, and Ficoll), 0.1% SDS, 100 µg/ml salmon sperm DNA, and 10⁶ cpm/ml radiolabeled probe. The filters were subsequently washed four times with 2× SSC/0.1% SDS and once with 0.2× SSC/0.1% SDS at 65 °C. Positive plaques were purified, and the cDNA inserts were subcloned into the EcoRI site of Bluescript vector. DNA sequences from both strands were determined by an automatic laser fluorescent sequencer (ABI).

RNA Analysis-- Total RNAs from different mouse tissues were prepared by the guanidinium-isothiocyanate extraction method (12) and separated on formaldehyde containing 1% agarose gel (12). After capillary transfer to nitrocellulose membranes, the blot was hybridized with the cDNA probe encompassing the entire mdGK-1 cDNA as described above. For RNase protection, the 1-185-nt region of mdGK-2 was amplified by the polymerase chain reaction and subcloned into Bluescript vector. Antisense riboprobe synthesis and RNase protection with mouse spleen poly(A) RNA was performed as described previously (13). The intensities of radioactive signals were quantitated by a phosphorimager (Bio-Rad GS525).

Expression, Purification, and Enzyme Assay of Recombinant dGK-- The coding region of the mouse and human dGKs was amplified using a set of primers containing NcoI (5' end) and BamHI (3' end) restriction sites. The polymerase chain reaction products were digested with NcoI and BamHI and subcloned into the pET-9d prokaryotic expression vector (Novagen), which provides a 6× His fusion tag under the control of the bacterial T7 promoter. Protein expression was induced in the Escherichia coli BL21 (DE3) pLysS strain by treatment with 1 mM isopropyl-1-thio--D-galactopyranoside for 3 h at 37 °C. The cells were lysed by sonication in a buffer containing 20 mM Tris-HCl, pH 8.0, 200 mM NaCl, 0.5% Triton X-100, 5 mM imidazole, and 1 mM phenylmethylsulfonyl fluoride, and the recombinant proteins were purified by affinity chromatography on a TALON (CLONTECH) metal affinity column as described previously (5). The eluted proteins were dialyzed in a buffer containing 50 mM Tris-HCl, pH 7.5, 10 mM NaF, 2 mM dithiothreitol, 20% glycerol, and 1 mM phenylmethylsulfonyl fluoride and used for SDS-polyacrylamide gel electrophoresis analysis (12) and enzyme assays.

Enzyme Assays-- The activities of recombinant mouse and human dGK were compared by phosphoryl transfer assay using [³²P]ATP as the phosphate donor and various nucleosides as phosphate acceptors. Reaction mixtures contained 50 mM Tris-HCl, pH 7.6, 5 mM MgCl₂, 5 mM dithiothreitol, 0.5 mg/ml bovine serum albumin, 15 mM NaF, 0.4 mM [³²P]ATP, 7.5 ng of purified enzyme, and 5 or 100 µM nucleoside substrate. Reaction mixtures were incubated at 37 °C for 20 min, and reactions were stopped by heating at 100 °C for 2 min. The phosphorylated products were separated by polyethyleneimine thin layer chromatography and quantified as described previously (14).

Detection of dGK-GFP Fusion Proteins in Mammalian Cells-- The coding regions of mdGK-1 and mdGK-2 were subcloned into the NheI and XhoI site of the pEGFP-N1 (CLONTECH) vector to generate in-frame carboxyl-terminal fusions with GFP cDNA under the control of the cytomegalovirus promoter. The resulting constructs were used for transient transfections of Cos-1 cells by the calcium phosphate coprecipitation method (15). 24 h after transfection, the intracellular localization of GFP fusion proteins in live cells was observed by fluorescence microscopy in a Leitz Dialux 20 EB microscope equipped with epifluorescence optics (15).

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

We have screened a mouse spleen cDNA library with the human dGK cDNA probe and isolated six positive clones. Sequence analysis revealed two types of clones, mdGK-1 and mdGK-2. The mdGK-1 cDNA sequence contained a 831-nt-long open reading frame encoding a protein of 277 amino acids with a predicted molecular mass of 32 kDa (Fig. 1). The amino acid sequence exhibited a 75% identity to human dGK, with 58 conservative and 11 nonconservative amino acid substitutions throughout the coding region. Amino acid changes were most frequent at the amino-terminal region containing the mitochondrial translocation signal sequence (20 substitutions out of 39 amino acids), whereas certain domains such as the putative ATP binding site, the substrate specificity site, and the arginine-rich domain were almost identical (Fig. 1). These motifs are highly conserved in several deoxynucleoside kinases of different species (16), underlining their functional significance.

View larger version (44K): [in this window] [in a new window]   Fig. 1.   The nucleotide and deduced amino acid sequence of mouse dGK-1 cDNA. Amino acids in bold indicate altered residues as compared with the human dGK sequence (5, 6). The highly conserved regions in different viral and mammalian deoxynucleoside kinases are boxed. The black arrow indicates the potential mitochondrial peptide cleavage site.

The nucleotide sequence of the mdGK-2 clone was identical to mdGK-1 cDNA from nucleotide position 43 (which corresponds to nt 14 of mdGK-1) throughout the rest of its coding and 3'-untranslated region (Fig. 2). However, the sequence divergence in the 5' end region eliminates the first ATG codon found in mdGK-1, generating an open reading frame starting from a downstream ATG codon that corresponds to amino acid position 32 of mdGK-1. The two in-frame stop codons in the divergent part of mdGK-2 exclude the possibility of potential start sites upstream of the cloned cDNA (Fig. 2). mdGK-2 cDNA therefore encodes an amino-terminally truncated isoform of dGK, lacking the first 31 amino acids of the putative mitochondrial signal sequence. Although it contains some of the amino acids of the consensus mitochondrial cleavage site (Fig. 2), the characteristic feature of most mitochondrial targeting sequences, an upstream -helical region followed by a stretch of positively charged amino acids (17, 18), is absent. This raised the possibility that mdGK-2 may encode a cytoplasmic isoform of dGK. To confirm this possibility, we investigated the subcellular localization of mdGK-1 and mdGK-2. Two plasmids were constructed to express mdGK-1 and mdGK-2 as fusion proteins with the GFP. The GFP moiety was inserted into the carboxyl-terminal part to avoid interference with the amino-terminal signal sequence. Fluorescence microscopy of transiently transfected Cos-1 cells revealed dramatic differences in the localization of the two fusion proteins. Transfection of mdGK-1-GFP gave rise to several dot-like fluorescent structures in a nonfluorescent cytoplasmic background (Fig. 3), characteristic of the mitochondrial location of proteins (19, 20). In contrast, transfection of mdGK-2-GFP produced a diffused fluorescent signal throughout the cytoplasm (Fig. 3). In both cases, no fluorescence was detected in the nucleus. These data strongly suggest that mdGK-2, even when overexpressed, does not translocate into the mitochondria.

View larger version (16K): [in this window] [in a new window]   Fig. 2.   Sequence comparison of the amino-terminal region of mdGK-1 and mdGK-2. The nucleotide sequence of mdGK-2 cDNA is identical to mdGK-1 from nt position 43. The divergent upstream sequence eliminates the first ATG codon of mdGK-1, placing the translation initiation site at the downstream ATG at nt position 130. Asterisks indicate in-frame stop codons in the 5' end of mdGK-2. The presumed consensus motif and the site for mitochondial cleavage (18) are indicated by bold letters and an arrow, respectively.

View larger version (20K): [in this window] [in a new window]   Fig. 3.   Analysis of intracellular localization of mdGK-1 and mdGK-2. Fluorescent images of live Cos-1 cells transfected with pEGFP-N1 (GFP), pEGFP-N1-mdGK-1 (mdGK-1-GFP), and pEGFP-N-1-mdGK-2 (mdGK-2-GFP) are shown. The expected diffuse fluorescent signal in all cellular compartments is seen with GFP alone, whereas mdGK-1-GFP and mdGK-2-GFP show patterns characteristic of mitochondrial and cytoplasmic proteins, respectively.

We next examined the expression pattern of mouse dGK in different tissues. Northern blot analysis of total RNAs with the coding region of mdGK-1 showed a broad band around 1.3 kb in all tissues tested (Fig. 4). However, the relative strength of the signals varied greatly, suggesting clear tissue-specific differences in dGK expression. The highest levels were observed in the spleen and thymus, whereas much lower levels of dGK mRNA were detected in the brain and liver (Fig. 4).

View larger version (45K): [in this window] [in a new window]   Fig. 4.   Tissue-specific expression of mouse dGK. Blots containing 20 µg of total RNAs from the indicated mouse tissues were hybridized with the cDNA probes of mdGK-1 (top panel) and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as control. Numbers on the right indicate the positions of the molecular size marker in kb.

To determine the expression levels of mdGK-2 relative to mdGK-1, we performed a RNase protection experiment using highly purified poly(A) mRNA from mouse spleen, where dGK mRNA was most abundant. A 194-nt-long antisense riboprobe containing the 1-185-bp region of mdGK-2 cDNA was synthesized, which should give rise to a 185-bp protected fragment when hybridized to mdGK-2 mRNA and to a 142-bp fragment when hybridized to mdGK-1 mRNA. The results of this assay demonstrate a substantial expression of the mdGK-2 isoform, which accounts for about 14% of the total dGK (mdGK-1 + mdGK-2) mRNA population (Fig. 5).

View larger version (25K): [in this window] [in a new window]   Fig. 5.   Determination of the relative abundance of dGK isoforms in the mouse spleen. A RNase protection experiment was performed using a 194-nt-long riboprobe synthesized from p-Bluescript-mdGK2 (1-185 nt) plasmid and mouse spleen poly(A) RNA. Lane 1, probe containing extra sequences from the vector. Lane 2, mouse spleen poly(A) RNA. Lane 3, yeast tRNA control. The intensities of the 185-bp and 142-bp protected fragments that correspond to mdGK-2 and mdGK-1 were quantitated by phosphorimager analysis. The values obtained were normalized according to the size and U content of each fragment and expressed as a percentage of the total mdGK RNA population (numbers at the bottom).

For biochemical characterization of mouse dGK, we expressed its cDNA as well as its human counterpart in E. coli as 6× His-tagged fusion proteins. The histidine tag allowed a one-step affinity purification of the recombinant proteins, giving highly pure products of ~33 kDa as judged by SDS-gel electrophoresis (data not shown). The phosphoryl transfer assay was used to compare the activities of the human and mouse dGK using various substrates. The substrate specificity of mouse dGK was similar to that of the human enzyme (Table I). Both enzymes phosphorylate deoxyguanosine, deoxyadenosine, and their analogues at both low (5 µM) and high (100 µM) concentrations. The phosphorylation of the cytostatic nucleoside analogues 2-chloro-2-deoxyadenosine, 2',2'-difluorodeoxyguanosine, and 2'-2-fluoro-9--D-arabinofuranosyladenine was achieved to about the same extent with both enzymes. The anti-human immunodeficiency virus compound 2',3'-dideoxyinosine was also phosphorylated to about 10% of that of deoxyguanosine at a physiological concentration. The major difference observed here was that the human enzyme showed a greater affinity toward deoxyadenosine and was able to phosphorylate deoxycytidine at a high substrate concentration. Both enzyme preparations showed specific activities similar to that reported previously for the human dGK enzyme (5).

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

An important property of all four mammalian deoxynucleoside kinases, especially in the case of the enzyme family comprising dCK, dGK, and TK2, is their broad substrate specificity (16). This enabled the development of a large number of anti-cancer and anti-viral nucleoside analogues whose metabolism to their active cytotoxic forms depends on the activity of these enzymes. The need for the establishment of relevant animal models is enormous because most of our knowledge of dNTP metabolic pathways and their pharmacological exploitation comes from studies using immortalized cell lines. A prerequisite of such models is to compare the main biochemical properties of the enzymes involved between different species. Human and mouse TK1 or TK2 appear to have similar substrate specificities against several deoxyribonucleoside analogues (21). In contrast, significant differences in the capacity to phosphorylate a number of substrates were observed between human and mouse dCK (21, 22).

In the present study, we have cloned and analyzed the mouse homologue of dGK as an initial step to compare its biochemical properties and contribution to dNTP metabolism with those of its human counterpart. The amino acid sequence of mouse dGK showed a 75% identity to human dGK, significantly lower than the identity observed between mouse and human TK1 (87%) (23) or dCK (94%) (22). This suggests a lower selective pressure on the dGK coding region that may relate to its functional redundancy. On the other hand, all motifs initially identified in herpesvirus thymidine kinases and subsequently found in dCK and TK2 (24, 25) are highly conserved. These include the putative ATP binding site, the substrate recognition site, and the arginine-rich site (Fig. 1). Therefore, the substrate preferences of deoxynucleoside kinases should depend on their different active conformations, determined in part by other, less conserved regions. We found similar specificities between mouse and human dGK for most of the substrates studied. This information, together with careful comparative sequence analyses of the less conserved regions, will help to design domain swapping and site-directed mutagenesis experiments to elucidate the structural basis of the overlapping but not identical substrate specificities of these enzymes.

The high variations of dGK mRNA in different mouse tissues suggest different cell type-specific requirements for this enzyme. dGK mRNA levels were highest in mouse spleen and thymus and were very low in the brain and liver. This pattern is similar to the previously reported expression pattern of mouse dCK mRNA, which was shown to be most abundant in lymphoid tissues and relatively rare in other tissues (22). This finding suggests that the expression of the two enzymes is not regulated to compensate for each other but rather to provide increased rates of salvage biosynthesis in certain cell types.

dGK has been considered to be located exclusively in the mitochondria. This assumption is based on two important findings: (a) the highest dGK enzyme activity was found in purified mitochondrial extracts (26), and (b) human dGK cDNA contained a functionally active mitochondrial targeting sequence (6, 27). In this study, we found that mouse spleen cells express an unusual mRNA species of dGK (mdGK-2) whose open reading frame codes for an amino-terminally truncated, active isoform that lacks a mitochondrial signal peptide. The perfect identity with mdGK-1 in the rest of the sequence corroborates the idea that this isoform is probably generated by an alternative splicing mechanism. When expressed as a GFP fusion protein, mdGK-2 cannot enter the mitochondria, suggesting that it represents a cytoplasmic isoform of dGK.

Although the majority of the mouse spleen dGK mRNA population (86%) encodes for the characteristic mitochondrial isoform (mdGK-1), the amount (14%) of the truncated isoform coding mRNA (mdGK-2) is high enough to suggest a significant functional importance. It is interesting to note that the cDNA isolated for the other mitochondrial kinase, human TK2, also lacks a mitochondrial signal sequence (7, 8). Although the structure of full-length TK2 cDNA remains to be determined, the reported TK2 cDNAs may also represent true non-mitochondrial isoforms rather than simply corresponding to partial cDNA clones. The expression of a cytoplasmic dGK raises several concerns with respect to our view of dNTP metabolism. It may participate in the supply of purine deoxynucleotides for nuclear DNA replication and repair. In this way, nucleoside analogues phosphorylated by dGK may exert their cytotoxic effects by interference with both mitochondrial and nuclear DNA synthesis. The location of dCK (28) and some of dGK in the cytoplasm also raises speculations of potential heterodimer formation between the two proteins that may generate novel enzyme properties. The recent identification of a functional mitochondrial dCTP carrier protein adds a further level of complexity to cellular dNTP metabolism (29). The activity of this carrier protein may make cytoplasmic pools of dNTPs available to mitochondria. This possibility is further strengthened by the mitochondrial toxicity observed with dideoxycytidine, a nucleoside analogue phosphorylated by cytoplasmic dCK (30). It seems that mammalian cells acquired several interdependent mechanisms that involve multiple redundant activities for both nuclear and mitochondrial dNTP supplies. The actual abundance and activities of these enzymes in distinct cellular compartments may thus provide important regulatory means for the cell to accommodate different physiological requirements. Given the pharmacological importance of cellular deoxynucleoside kinases, additional studies aimed at the elucidation of the regulatory mechanisms that control their cell type-specific expression and subcellular location-dependent contribution to different metabolic pathways are clearly warranted.

	FOOTNOTES

^* This work was supported in part by European Union Grant BIOMED BMH4-CT-96-0479 and the Greek General Secretariat for Science and Technology.The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

The nucleotide sequence(s) reported in this paper has been submitted to the GenBank^TM/EMBL Data Bank with accession number(s) AJ133749 and AJ133750.

^§ Supported by the Joint Graduate Program in Molecular Biology and Biomedicine of the Greek Ministry of Education.

To whom correspondence should be addressed. Tel.: 30-81-391173; Fax: 30-81-391101; E-mail: talianid@nefeli.imbb.forth.gr.

	ABBREVIATIONS

The abbreviations used are: dNTP, deoxynucleotide triphosphate; dCK, deoxycytidine kinase; dGK, deoxyguanosine kinase; mdGK, murine dGK; GFP, green fluorescent protein; TK, thymidine kinase; nt, nucleotide(s); bp, base pair(s).

	REFERENCES

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

CiteULike

Complore

Connotea

Del.icio.us

Digg

Reddit

Technorati

This article has been cited by other articles:

				W. Lewis Defective mitochondrial DNA replication and NRTIs: pathophysiological implications in AIDS cardiomyopathy Am J Physiol Heart Circ Physiol, January 1, 2003; 284(1): H1 - H9. [Full Text] [PDF]

				A. H. Sjöberg, L. Wang, and S. Eriksson Antiviral Guanosine Analogs as Substrates for Deoxyguanosine Kinase: Implications for Chemotherapy Antimicrob. Agents Chemother., March 1, 2001; 45(3): 739 - 742. [Abstract] [Full Text]

				M. Jullig and S. Eriksson Apoptosis Induces Efflux of the Mitochondrial Matrix Enzyme Deoxyguanosine Kinase J. Biol. Chem., June 22, 2001; 276(26): 24000 - 24004. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Petrakis, T. G. Articles by Talianidis, I. Search for Related Content PubMed PubMed Citation Articles by Petrakis, T. G. Articles by Talianidis, I. Social Bookmarking                      What's this?