Mutagenesis Advance Access originally published online on July 12, 2005
Mutagenesis 2005 20(5):345-350; doi:10.1093/mutage/gei045
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Modified bacterial mutation test procedures for evaluation of peptides and amino acid-containing material
Charles River Laboratories, Preclinical Services, CTBR, 87 Senneville Road, Senneville, Quebec, Canada H9X 3R3, 1Zelos Therapeutics, Building M-54, 1200 Montreal Road, Ottawa, Ontario, Canada K1A 0R6 and 2Covance Laboratories Ltd, Otley Road, Harrogate, North Yorkshire HG3 1PY, UK
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Biological materials can release amino acids during the course of bacterial mutation testing. Low levels of released amino acids from soluble materials can cause moderate increases in the number of revertant colonies on the plate, whereas higher levels lead to overgrowth of the background lawn, making counting of revertant colonies impossible. For poorly soluble material, the released amino acids can be present at high levels in localized spots on the plate, leading to the growth of ‘pseudorevertant’ colonies. The ‘treat and wash’ modified preincubation method employed here is an adaptation of the treat and plate method (used for evaluation of antibiotics) and involves washing the bacteria free of test compound after a 90 min exposure prior to plating out on minimal plates. The MC overlay method is a modified version of the standard plate incorporation assay, in which a top overlay containing 4% high viscosity methylcellulose is used in place of agar to stabilize the test compound in solution, preventing precipitation and subsequent localized amino acid release. Both modified methods produce the expected results for negative and positive controls. Peptides [synthetic curtailed analogs of human parathyroid hormone, PTH(1–34) and Ostabolin-CTM] that produced false positive results or could not be evaluated owing to overgrowth of the background lawn using standard methods, showed no artifacts and no evidence of genotoxicity using the modified methods. It is concluded that the treat and wash and MC overlay methods are valid versions of the bacterial mutation test for avoiding complications associated with released amino acids.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
The International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) consists of a series of committees, mainly concerned with standardizing the range of toxicology tests needed for new pharmaceuticals. ICH S2B guidance was finalized in 1997 and adopted by the US, European and Japanese regulatory authorities shortly after. The guidance indicates that the expected standard battery of genotoxicity tests normally required for small molecule pharmaceuticals should consist of a bacterial mutation test, an in vitro chromosome aberration test or a mouse lymphoma assay, and a rodent bone marrow micronucleus test. ICH document S6 (1997) gives additional guidance on testing of biological molecules including hormones, growth factors and proteins. ICH S6 states that genotoxicity studies normally conducted for pharmaceuticals are not applicable or needed for biologicals, such as peptides, which are considered unlikely to react directly with DNA. It indicates that genotoxicity studies may be required on a case-by-case basis, e.g. where a chemical linking agent is used to modify the molecule or when potentially mutagenic impurities or degradants are present. Although the guidance indicates that the standard genotoxicity studies are inappropriate for assessing biologicals in such cases, it does not give specific guidance on which studies would be appropriate.
In practice, the bacterial mutation test is the most widely used and best validated bioassay for assessment of potentially mutagenic contaminants. Since it is relatively inexpensive and sensitive, it seems an appropriate assay to use to confirm the absence of genotoxic contaminants produced during synthesis, or remaining after purification of biological molecules (1
–3).
Although the bacterial mutation assay may be the first choice for routine genetic toxicity screening of chemicals, it has long been recognized that materials containing, or capable of releasing amino acids can interfere with the assay. For example, biological samples (urine, feces, food), proteins, peptides and histidine itself can cause additional growth of Salmonella on minimal medium plates and result in additional spontaneous mutations and/or overgrowth of the background lawn of non-revertant bacteria which obscures revertant colonies (4–9).
To avoid misinterpretations when testing histidine/tryptophan containing compounds, it has been proposed that a modified preincubation method with extensive washing prior to plating could be employed (4,10). Although various complex washing methods have been demonstrated by other groups (11–13), the present study verifies that a simple modification of the ‘treat and plate’ method (14), i.e. washing the bacteria free of test compound after a 90 min exposure prior to plating out on minimal plates (‘treat and wash’ method), avoids false positive responses and overgrowth of the background bacterial lawn resulting from free amino acids, while producing clear positive responses with appropriate positive controls.
Precipitation of a peptide in the top agar with subsequent release of amino acids can also lead to confounding results. A modification of the standard plate incorporation method (‘MC overlay’ method) is presented which has been used to stabilize a peptide in solution and avoid confounding artifacts.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Chemicals
9-Aminoacridine (9AC), 2-aminoanthracene (2AA), benzo[a]pyrene (BaP), histidine, methylcellulose (MC), reagent or pharmaceutical grade is acceptable, 2-nitrofluorene (2NF), 4-nitroquinoline N-oxide (NQO), sodium azide (NaAz) and tryptophan were obtained from Sigma-Aldrich (Oakville, Canada). 9AC, 2AA, BaP, 2NF and NQO were formulated in DMSO; histidine, MC, tryptophan and NaAz were formulated in water.
Peptides
PTH(1–34) and Ostabolin-CTM were independently developed as stable synthetic curtailed analogs of human parathyroid hormone for potential treatment of osteoporosis.
Peptide PTH(1–34) corresponds to the first 34 amino acids of the human parathyroid hormone. The peptide was manufactured by Bachem, Torrance, CA, USA (86.84% pure; all concentrations are expressed in terms of active ingredient). PTH(1–34) has a molecular weight of 4118 Da, and contains one tryptophan and three histidine residues in the sequence H2N-Ser-Val-Ser-Glu-Ile-Gln-Leu-Met-His-Asn-Leu-Gly-Lys-His-Leu-Asn-Ser-Met-Glu-Arg-Val-Glu-Trp-Leu-Arg-Lys-Lys-Leu-Gln-Asp-Val-His-Asn-Phe-OH. PTH(1–34) was formulated as a solution in water on each occasion of use, just prior to use.
Ostabolin-CTM corresponds to the first 31 amino acids of the human parathyroid hormone with a Leu27, cyclo(Glu22-Lys26) modification (15). The peptide was supplied by Zelos Therapeutics, Ottawa, Canada (
85% pure; all concentrations expressed as material as supplied). The peptide has a molecular weight of 3685 Da, and contains one tryptophan and two histidine residues in the sequence H-Ser-Val-Ser-Glu-Ile-Gln-Leu-Met-His-Asn-Leu-Gly-Lys-His-Leu-Asn-Ser-Met-Glu-Arg-Val-Glu-Trp-Leu-Arg-Lys-Leu-Leu-Gln-Asp-Val-NH2. Ostabolin-CTM has limited solubility in water; therefore, it was formulated as an apparent solution in aqueous 1% (w/v) MC (400 cps) on each occasion of use, just prior to use.
Bacterial strains
Salmonella typhimurium strains TA1535, TA1537, TA98 and TA100 and Escherichia coli strain WP2 uvrA were originally supplied by Moltox (NC, USA) and were characterized and maintained as previously recommended (14,16).
S9 mix
Phenobarbital/5,6-benzoflavone-induced male Sprague–Dawley rat liver fraction (S9 fraction) was supplied by Moltox (NC, USA). S9 mix contained 10% (by volume) S9 fraction unless otherwise indicated and the following sterile co-factors: 8 mM magnesium chloride, 33 mM potassium chloride, 100 mM sodium phosphate buffer pH 7.4, 5 mM glucose-6-phosphate and 4 mM nicotinamide adenine diphosphate.
Bacterial mutation tests
Tests were based on previously recommended procedures (16). In addition to histidine and biotin, the top agar contained minimal tryptophan (0.05 mM) to support E.coli growth (17). Histidine, tryptophan and synthetic PTH(1–34) were evaluated in the standard plate incorporation, pre-incubation and treat and wash assays. Ostabolin-CTM was evaluated in the standard plate incorporation and MC overlay assays.
Standard plate incorporation assay
A 0.5 ml aliquot of S9 mix or phosphate buffer 0.2 M pH 7.4 was combined with 0.1 ml late log bacterial culture in a sterile container. A 0.1 ml aliquot of the treatment solution was added, and 2 ml of molten top agar was added immediately afterward. The solution was overlaid onto a minimal glucose plate (1.5% agar, Vogel–Bonner medium E, 2% glucose). The plates were inverted and incubated at 37°C for 48 to 72 h.
Preincubation assay
Procedures were as per the standard plate incorporation assay with the exception that the S9 mix/buffer, bacteria and treatment were incubated for 30 min with shaking (180 r.p.m.) at 37°C before the addition of molten top agar.
Treat and wash assay
Procedures were as per the preincubation assay with the exception that the preincubation time was increased from 30 to 90 min. The extended duration of bacterial exposure during the preincubation compensated for the absence of bacterial exposure on plates (as the test compound was washed away prior to plating). After the 90 min preincubation, 15 ml of a wash solution of Oxoid No. 2 nutrient broth in phosphate buffered saline (1:7 v/v) was added and the washed bacteria were collected by centrifugation at 2000 g for 30 min. All but 
0.7 ml of the supernatant was removed and discarded, and the bacteria were resuspended in the residual supernatant prior to plating via top agar.
MC overlay assay
An MC overlay was prepared on the day of the test by combining 5 g sodium chloride and 40 g MC (4000 cps) in 900 ml of water heated to 80°C. The suspension was kept stirring and cooled to 50°C. A solution of histidine, tryptophan and biotin (each 0.5 mM) was added (100 ml per 900 ml sodium chloride/MC) and the mixture was stirred at 50–60°C throughout use. A 0.5 ml aliquot of S9 mix or phosphate buffer 0.2 M pH 7.4 was combined with 0.1 ml overnight bacterial culture in a sterile container. A 2 ml aliquot of the MC overlay suspension was added to the tube, and a 0.1 ml aliquot of the treatment solution was added immediately afterward. The mixture was overlaid on a prewarmed (37°C) minimal glucose plate. The plates were held at 4°C for 1 h after plating to ensure gelling of the MC overlay. The plates were incubated agar side down (not inverted) to maintain the position of the MC layer. Typically, plates are inverted during incubation to prevent condensation droplets from falling onto the surface of the agar. Incubating the MC overlay plates agar side down has the potential drawback of condensation droplets causing the spread of colonies resulting in artificially high colony counts or indistinguishable colonies (unscorable plate); however, this was not observed in our laboratory.
Evaluation and interpretation of results
Sterility, negative/vehicle and positive controls were included in each experiment. Triplicate plates were evaluated at each experimental point. Revertant colonies were routinely counted using an automated colony counter (ProtoCOL by Synbiosis, Cambridge, UK). If precipitate interfered with the operation of the system, counts were enumerated manually. The background lawn was evaluated using an inverted microscope. Toxicity was identified by a substantial reduction in the integrity of the lawn, or a marked reduction in the number of revertants compared with the vehicle control (fold response <0.6). A positive mutagenic response was defined as a dose-related increase in revertant colony numbers to at least twice the concurrent untreated control levels (1.5x for strain TA100).
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Treat and wash assay
The treat and wash bacterial mutation test was employed to evaluate the mutagenicity of soluble biological material capable of releasing histidine and tryptophan. Histidine and tryptophan can cause overgrown background bacterial lawns and variable/sporadic increases in revertants in the plate incorporation and preincubation versions of the bacterial mutation test (Table I). Modification of the procedures to include a longer preincubation followed by a washing step to remove the amino acids prior to plating (‘treat and wash’ method) prevented overgrown lawns and significant increases in revertant colony counts (Table II).
|
|
A soluble peptide, PTH(1–34), caused overgrowth of the background bacterial lawn for all strains at levels of
500 µg/plate in the presence of S9 using the plate incorporation method, presumably as a result of S9-mediated release of histidine and tryptophan (Table III). The overgrown lawns prevented any assessment of revertant colony counts. Using the treat and wash method in the presence of S9 in a confirmatory assay, no overgrowth of the background lawn was observed for any strain at any concentration of the peptide (Table III). Thus, revertant colony counts were available for the peptide up to the highest dose level recommended by regulatory guidelines (5000 µg/plate). PTH(1–34) did not cause any substantial increases in revertant colony counts either in the absence or in the presence of S9 mix using plate incorporation, preincubation or treat and wash methods.
|
MC overlay assay
The development of the MC overlay bacterial mutation test was required because of problems in assessing a peptide, Ostabolin-CTM, that was not freely soluble at the higher concentrations typically used for bacterial mutation testing (i.e. levels up to 5000 µg/plate).
Ostabolin-CTM caused apparent increases in colony numbers of strains TA98 and WP2 uvrA in the presence of S9 using the plate incorporation method (occasion 1 in Table IV). No increases in colony counts were observed with any other strain in the presence or absence of S9 (data not shown). Precipitate was observed on the plates at the highest dose level evaluated (5000 µg/plate), which, in some cases, prevented the evaluation of the background lawn. Two subsequent plate incorporation assays were performed: one with TA98 and WP2 uvrA in the presence of S9 only (occasion 2), and the other with all the strains in the absence and presence of S9 mix containing increased S9 fraction, i.e. 30% per volume (occasion 3). Increased colony counts were again obtained in TA98 and WP2 uvrA (occasion 2 in Table IV), and in TA1535, TA98 and WP2 uvrA (occasion 3 in Table IV). No increases in colony counts were observed with any other strain in the presence or absence of S9 for occasion 3 (data not shown). Precipitate was observed at concentrations 500 µg/plate on occasions 2 and 3.
|
To ensure that the response was not the result of a mutagenic impurity in the peptide, Ostabolin-CTM was re-purified by the manufacturer and tested using strains TA98 and WP2 uvrA in the presence of S9 using the plate incorporation method. Increases in colony counts were obtained for strain TA98 but not WP2 uvrA (data shown for TA98 as occasion 4 in Table IV). No precipitate was observed. Subsequently, a new batch of the peptide was tested under similar conditions using all five standard tester strains. On this occasion, Ostabolin-CTM caused an increase in colony counts in strain TA1535 only (results for TA1535 shown as occasion 5 in Table IV). Precipitate was observed at all doses evaluated (
500 µg/plate).
Although the genotype of the apparent revertant colonies was not verified using replica plating (18), the bacteria were clearly distinguishable from precipitation using an inverted microscope. At this point, it was noted that there did seem to be some association between bacterial colonies and the precipitate, although the standard sterility check showed that the bacteria did not originate from the test article.
Finally, the bacterial mutation test was modified to use an overlay formulated with high viscosity MC as the gelling agent instead of agar. Using this MC overlay method, no precipitation was observed and no substantial increases in revertant colony counts were obtained at any dose level of the peptide with any strain either in the absence (data not shown) or presence of S9 mix containing 10% or 30% S9 fraction by volume (Table V).
|
For all assays, the negative/vehicle control values were within the laboratory historical range, the positive controls induced substantial increases in revertant colony counts and no toxicity was identified by a substantial reduction in the integrity of the lawn confirming the sensitivity of the bacterial strains, the activity of the S9 and the absence of any adverse affects on the bacteria and S9 by the treat and wash or MC overlay methodology.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Treat and wash assay
The treat and wash method described in the present study is based on the treat and plate method (14
) sometimes used to assess the mutagenicity of antimicrobial agents and recommended when feeding effects are suspected (19). Since peptides are expected to be of relatively low toxicity, no assessment of survivability is required, which considerably decreases the amount of work involved. The same criteria are employed for assessment of mutagenicity as in a standard bacterial mutation test, i.e. the observed increase in the absolute number of revertants per plate, rather than through an indirect calculation of mutation frequency based on colony counts and viability following treatment. Therefore, calculation of mutation frequency and resultant introduction of a high degree of variance (owing to the practicalities of assessment of viability) is avoided. In the modified method, the bacteria are diluted/washed with phosphate buffered saline supplemented with nutrient broth after the preincubation. In order to conserve viability at plating, it was essential to supplement the wash buffer with nutrient broth (to mimic the presence of minimal nutrient broth at the time of plating in a standard test).
The dosage and particular positive controls employed were modified (where necessary) to achieve a clear positive result using the treat and wash method; however, all the positive responses were reproducible. Although the treat and wash method shows increased sensitivity to 9AC and NQO (presumably as a result of the longer preincubation period), the method seems to show slightly reduced sensitivity toward other mutagens (e.g. NaAz, 2AA and BaP). The reduced sensitivity is probably the result of the limited duration of exposure and because the bacteria are not rapidly dividing during the exposure phase. Similar findings were obtained with a related miniaturized preincubation method (20). It may be that the treat and wash method requires further supplementation to allow cell division during the bacterial exposure period and enhance sensitivity to these mutagens. Even so, using the current methodology, the treat and wash method did allow evaluation of a soluble peptide up to the standard maximum of 5000 µg/plate in the presence of S9 mix without any overgrowth of the background bacterial lawn, which is 30 times higher than the maximum level that could be evaluated using standard methodology.
MC overlay assay
The sporadic positive responses and variable precipitate observed in the presence of S9 led to the proposal that precipitated Ostabolin-CTM was being hydrolyzed by enzymes in the S9 mix, causing a localized release of amino acids around the particulate material and an increase in bacterial growth/colony counts (‘pseudorevertants’) using the standard plate incorporation method. Substitution of MC for agar in the overlay stabilized the peptide in solution; so, no precipitate was observed, and allowed evaluation of Ostabolin-CTM up to the maximum dose recommended by regulatory guidelines (5000 µg/plate) in the absence of S9 mix and in the presence of S9 mix containing 10 and 30% S9 fraction by volume. The fact that the negative and positive controls produced expected responses, and no substantial reductions in the integrity of the lawns was observed, demonstrates that the concentration of MC used in the overlay was non-toxic, the MC had no adverse effect on the bacteria or S9 mix. The actual nature of the pseudorevertant colonies obtained using standard methodology is not known, i.e. they may have been composed of auxotrophic bacteria, or a mixture of auxotrophic and wild-type revertants. The variable nature of the increases in colonies is explained by the nature of the precipitate which varied in density and particulate size between plates and especially between experiments.
The treat and wash and MC overlay versions of the bacterial mutation test are valid alternatives for avoiding complications associated with feeding effects owing to released amino acids. The presented minor modifications to the standard bacterial mutation test provide a rapid, inexpensive and sensitive solution to the ICH requirement of genotoxicity evaluation of biologicals, such as peptides, where the standard genetox test battery does not apply.
Conflict of interest statement
Crista Thompson works for a contract research organisation (Charles River Laboratories-CTBR) that was hired by two pharmaceutical companies (one being Zelos) to evaluate the genotoxicity of two peptides [Ostabolin-C and PTH(1–34)]. Paul Morley is the Scientific Founder and Chief Technical Officer of Zelos Therapeutics Inc., which is developing Ostabolin-C. The other authors have no conflicts to declare.
|
Acknowledgments |
|---|
The authors wish to thank Dr Bernard Rushton for many helpful discussions throughout this project and the Genetic Toxicology staff at Charles River Laboratories, Preclinical Services, CTBR for their expert technical assistance.
|
Notes |
|---|
* To whom correspondence should be addressed. Tel: +1 514 630 8254; Fax: +1 514 630 8230; Email: Ray.Proudlock{at}ca.crl.com
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
-
1. Cerna,M. (1985) Mutagenicity of beta-glucuronidase in the Ames test. Mutat. Res., 143, 169–172.[Web of Science][Medline]
2. Allen,J.S. and Panfili,J. (1986) Ames Salmonella/mammalian-microsome testing of peptides and peptide synthesis reagents. Mutat. Res., 170, 23–29.[CrossRef][Web of Science][Medline]
3. Castellino,S., de Castiglione,R., Forino,R., Galantino,M. and Pulci,R. (1991) Mutagenic contaminants in synthetic peptides obtained by an azide coupling. Mutagenesis, 6, 185–187.
4. Aeschbacher,H.U., Finot,P.A. and Wolleb,U. (1983) Interactions of histidine-containing test substances and extraction methods with the Ames mutagenicity test. Mutat. Res., 113, 103–116.[Web of Science][Medline]
5. Kazmer,S., Katz,M. and Weinstein,D. (1983) The effect of culture conditions and toxicity on the Ames Salmonella/microsome agar incorporation mutagenicity assay. Environ. Mutagen., 5, 541–551.[Web of Science][Medline]
6. Gruter,A., Friederich,U. and Wurgler,F.E. (1991) The mutagenicity of edible mushrooms in a histidine-independent bacterial test system. Fd. Chem. Toxic., 29, 159–165.[CrossRef][Web of Science][Medline]
7. Prival,M.J., Simmon,V.F. and Mortelmans,K.E. (1991) Bacterial mutagenicity testing of 49 food ingredients gives very few positive results. Mutat. Res., 260, 321–329.[CrossRef][Web of Science][Medline]
8. Albertini,S. and Gocke,E. (1993) Renin inhibitors as an example of presumptive irrelevant positive findings in the Salmonella/mammalian microsome assay (Ames test). Mutat. Res., 298, 237–246.[CrossRef][Web of Science][Medline]
9. Nylund,L. and Einisto,P. (1993) Mutagenicity testing of protein-containing and biological samples using the Ames/Salmonella plate incorporation test and the fluctuation test. Mutat. Res., 272, 205–214.[CrossRef]
10. Kirkland,D.J. and Kim,N.N. (1995) Special considerations for conducting genotoxicity tests with protein materials. Mutagenesis, 10, 393–398.
11. Mitchell,I.de.G., Dixon,P.A., Gilbert,P.J. and White,D.J. (1980) Mutagenicity of antibiotics in microbial assays: problems of evaluation. Mutat. Res., 79, 91–105.[CrossRef][Web of Science][Medline]
12. O'Connor,H.J., Axon,A.T.R., Riley,S.E. and Garner,R.C. (1984) Mutagenicity of gastric juice: the importance of controlling histidine concentration when using Salmonella tester strains. Carcinogenesis, 5, 853–856.
13. Verhagen,H., Bruijntjes-Rozier,G.C.D.M., Coenen,T.M.M. and Oosterom,J. (1994) Modified suspension Ames test for testing proteinaceous substances: an initial step. Fd. Chem. Toxic., 32, 1161–1166.[CrossRef][Web of Science][Medline]
14. Green,M.H.L. and Muriel,W.J. (1976) Mutagen testing using Trp+ reversion in Escherichia coli. Mutat. Res., 38, 3–32.[CrossRef][Web of Science][Medline]
15. Whitfield,J.F., Morley,P., Willick,G.E., Langille,R., Ross,V., MacLean,S. and Barbier,J.-R. (1997) Cyclization by a specific lactam linkage increases the ability of human parathyroid hormone (hPTH)-(1-31)NH2 to stimulate bone growth in ovariectomized rats. J. Bone Miner. Res., 12, 1246–1252.[CrossRef][Web of Science][Medline]
16. Maron,D.M. and Ames,B.N. (1983) Revised methods for the Salmonella mutagenicity test. Mutat. Res., 113, 173–215.[CrossRef][Web of Science][Medline]
17. Mortelmans,K. and Riccio,E.S. (2000) The bacterial tryptophan reverse mutation assay with Escherichia coli WP2. Mutat. Res., 455, 61–69.[Web of Science][Medline]
18. Pedersen,P., Thomsen,E. and Stern,R.M. (1983) Detection by replica plating of false revertant colonies induced in the Salmonella-mammalian microsome assay by hexavalent chromium. Environ. Health. Perspect., 51, 227–230.[Web of Science][Medline]
19. Gatehouse,D.G., Rowland,I.R., Wilcox,P., Callander,R.D. and Forster,R. (1990) Bacterial mutation assays. In Kirkland,D.J. (ed.), UKEMS Sub-committee on Guidelines for Mutagenicity Testing, Report, Part I revised, Basic Mutagenicity Tests: UKEMS Recommended Procedures. Cambridge University Press, Cambridge, pp. 13–61.
20. Kado,N.Y., Langley,D. and Eisenstadt,E. (1983) A simple modification of the Salmonella liquid-incubation assay. Increased sensitivity for detecting mutagens in human urine. Mutat. Res., 121, 25–32.[CrossRef][Web of Science][Medline]
Received on May 3, 2005; revised on June 10, 2005; accepted on June 13, 2005.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||