Analyst, December 1994, Vol. 119 2765

Enzyme lmmunoassay for the Detection of lsoxazolyl Penicillin Antibiotics in Milk*

Ewald Usleber, Marion Lorber, Margit Straka, Gerhard Terplan and Erwin Mart1 bauer Institute for Hygiene and Technology of Food of Animal Origin, Veterinary Faculty, University of Munich, Schellingstr. 10, 80799 Miinchen, Germany

Polyclonal antibodies were raised against isoxazolyl penicillins in rabbits after immunization with a cloxacillin-human serum albumin conjugate. The antisera were tested in direct and indirect competitive enzyme immunoassays (EIAs), using glucose oxidase or horseradish peroxidase conjugates of oxacillin, cloxacillin, or dicloxacillin, respectively, as the labelled antigen. The relative cross-reactivities of each test system with oxacillin, cloxacillin, and dicloxacillin, determined from the amount of antibiotic required for 50% inhibition of labelled antigen binding, were dependent on the antibiotic used as the labelled antigen. Other 6-lactam antibiotics did not cross-react in these test systems. In a direct EIA using a cloxacillin-horseradish peroxidase conjugate, cloxacillin and dicloxacillin in milk were detected at levels of 10 and 30 ng ml-1; the average recoveries at these levels were 102 and 84%, respectively.

Keywords: Oxacillin; cloxacillin; dicloxacillin; antibody specificity; milk

Introduction

Among the (3-lactam antibiotics, the penicillinase-stable isox- azolyl penicillins oxacillin, cloxacillin, and dicloxacillin (Fig. 1) are very efficient in the veterinary treatment of infections with (3-lactamase-producing pathogenic bacteria, such as some strains of Staphylococcus aureus, which is a major cause of bovine mastitis. In dry cows, isoxazolyl penicillins are predominantly used as long-acting intramammary depositing antibiotics, but intravenous or intramuscular injections are also given to treat systemic infections.' Prolonged drug excretion, unauthorized drug use, or neglect of the withdrawal times may result in residues of these substances in milk and tissues. Problems associated with antibiotic residues in milk include the risk of allergic reactions after consumption by penicillin-sensitized persons, increased resistance of pathogenic bacteria towards antibiotics, and inhibition of bacterial starter cultures used in dairy production.

Within the European Union (EU), maximum residue limits (MRLs) have been set for isoxazolyl penicillins in milk (30 pg kg-1) and animal tissues (300 pg kg-1).* In the USA, the tolerance level for cloxacillin residues in milk is 10 pg kg-1.3 The most common screening methods for antimicrobial drug residues are microbiological tests, based on the growth inhibition of a test micro-organism (e.g., B. stearothermo- philus var. calidolactis) .4 These tests have in general sufficient sensitivity for penicillin antibiotics. However, they lack specificity and do not allow the identification of individual substances as required by MRL regulations. Physico-chemical methods"." €or the detection of cloxacillin and dicloxacillin are

* Presented at The Second International Symposium on Hormone and Veterinary Drug Residue Analysis, Oud St-Jan, Bruges, Belgium, May 31-June 3 . 1994.

both expensive and time-consuming, and thus are restricted to confirmatory analysis.

Immunochemical approaches have found increasing im- portance in antimicrobial drug control, in particular as routine screening tests.7 Enzyme immunoassay (EIA) development for (3-lactam antibiotics in milk has so far been focused on penicillin G .s A commercial immunochemical(3-lactam group- specific test, that is also cross-reactive with cloxacillin, is available.9 However, production and characterization of antibodies exclusively reactive with isoxazolyl penicillins, and their use in EIA has so far (May 1994) not been reported. This paper describes the development and characterization of polyclonal antibody-based EIAs for these compounds, using homologous and heterologous labelled antigens10 in direct and indirect EIA format. An easy method for the detection of cloxacillin and dicloxacillin in milk samples, using a direct EIA technique, is described.

Experimental

Chemicals, Buffers, and Equipment

The (3-lactam antibiotics oxacillin, cloxacillin, dicloxacillin, ampicillin, amoxicillin, azlocillin, bacampicillin, carbencillin, cephalothin, epicillin, metampicillin, methicillin, moxalac- tam, penicillin G , penicillin V, phenethicillin, piperacillin, and ticarcillin were obtained from Sigma Chemicals (Deisen- hofen, Germany). Glucose Oxidase (GO), horseradish perox- idase (HRP), human serum albumin (HSA), goat anti-rabbit immunoglobulin G (IgG-HRP) conjugate, isobutylchlorofor- mate, triethylamine, N-hydroxysuccinimide, dicyclohexylcar- bodiimide (DCC), 3,3',5,5'-tetramethylbenzidine (TMB), and all other chemicals used were also from Sigma. Antibiotic standard solutions, in phosphate buffered saline (PBS; 0.01 mol I - * , pH 7.3; phosphate buffer containing 0.1 mol 1-1

NaCI) were prepared daily. For milk sample analysis, cloxacil- lin standards were prepared in distilled water containing 100 g 1-l of skim milk powder (Oxoid, Unipath, Hampshire, England). The dilution buffer for coating microtitre plates (Maxisorp, Nunc, Wiesbaden, Germany) with either antise- rum or G O conjugates was carbonate-bicarbonate buffer (0.05 mol 1-1, pH 9.6; 100 pl per well). To block the free protein binding sites of the plates, PBS containing 20 g I - ' of

0 CH3 0 CO,H

Fig. 1 R*=H. cloxacillin; R'=R2=CI. dicloxacillin.

Structure of isoxazolyl pcnicillins. RI=R*=H, oxacillin; R*=Cl,

Publ

ishe

d on

01

Janu

ary

1994

. Dow

nloa

ded

by U

nive

rsity

of

Win

dsor

on

27/1

0/20

14 0

4:32

:22.

View Article Online / Journal Homepage / Table of Contents for this issue

2766 Analyst, December 1994, Vol. 11 9

sodium caseinate (Sigma) was used (200 pl per well). The microtitre plate wash solution was distilled water containing 8.5 g 1 - 1 of NaCl and 0.25 ml 1-1 of Tween 20. Enzyme conjugate dilution buffer was PBS containing 10 g 1-1 of sodium caseinate. Enzyme substrate solution consisted of 0.2 rnol 1 - 1 potassium citrate buffer (pH 3.9) containing 0.003 rnol 1-1 H202 and 0.001 rnol 1-1 TMB (100 pl per well). The enzyme reaction stopping solution was 1 rnol 1-1 H2S04 (100 p1 per well). Colour development for the EIA reaction was measured with an AT 400 microtitre plate reader (SLT, Crailsheim, Germany). The absorbance data were evaluated and the results calculated using EIA calculation software developed by Martlbauer. 11

Synthesis of Conjugates

The immunogen (cloxacillin-HSA) was prepared by a mixed anhydride method.12 Cloxacillin (7 mg) was dissolved with 0.5 ml of dimethylformamide (DMF) in a glass vial and kept at a temperature between -5 and -10 "C in a water-ethanol bath. Isobutylchloroformate (4 pl) and triethylamine (3 pl) were added and reacted for 10 min. The mixture was then added to 6 mg of HSA (dissolved with 1.5 ml distilled water + 1 ml of DMF) and stirred at -5 to -10 "C. The pH of the reaction mixture was frequently controlled and adjusted to approxi- mately 8 with carbonate-bicarbonate buffer (0.05 rnol I - l , pH 9.6). After a reaction time of 3 h, the cloxacillin-HSA conjugate was dialysed against 3 x 5 1 of PBS. The protein content of the solution, as determined by the method of Lowry et al. ,I3 was 1.9 mg ml-1. The conjugate was stored lyophilized at -20 "C.

The labelled antigens (HRP and GO conjugates) were prepared by an activated ester method.14 HRP conjugates of oxacillin, cloxacillin, and dicloxacillin, respectively, were prepared by reacting 21.8 pmol of antibiotic, 109 pmol of N-hydroxysuccinimide, and 218 pmol of dicyclohexylcarbodi- imide in 2 ml of DMF at room temperature for 16 h. Of this intermediate, 200 p1 were added to 0.218 pmol of HRP (in 2 ml of 0.13 rnol 1 - 1 NaHC03) and reacted for 2 h at room temperature, then the conjugate was dialysed against PBS (3 x 5 I). Precipitates were removed by centrifugation (10oOg; 10 min). The peroxidase concentrations of the supernatants, as determined spectrophotometrically at 403 nm, were 3.3 mg ml-1 (oxacillin-HRP), 3.4 mg ml-1 (cloxacillin-HRP), and 4.1 mg ml-1 (dicloxacillin-HRP). The conjugates were stored lyophilized at -20 "C. Glucose oxidase conjugates of the three antibiotics, for use in indirect EIAs, were prepared using the same conjugation procedure as for the HRP conjugates, except that the molar ratio of antibiotic to GO was 100 : 1 in the reaction mixture. The protein concentration13 of the conjugates was 3.2 mg ml-1 (oxacillin-GO), 2.7 mg ml-1 (cloxacillin-GO), and 3.3 mg ml-1 (dicloxacillin-GO), respectively.

Immunization and Antibody Titre Determination

For use as the immunogen, 500 pg of cloxacillin-HSA were emulsified in 2 ml of Freund's complete adjuvant and distilled water (3 + 1). Rabbits (three female chinchilla bastards, Savo Ivanova, Kisslegg, Germany) were each immunized with 2 ml portions of the emulsion by using multisite intradermal injections. Booster injections, using the same composition and amount of immunogen, were given subcutaneously 10 (rabbit 1) and 24 weeks (rabbits 1, 2, and 3) after the primary injection. Blood was collected from the Arteria auricularis magna. The relative antibody titre was determined in a double antibody solid-phase EIA as described by Martlbauer et al. ,15

using cloxacillin-HRP at a concentration of 1.7 pg ml-1. The titre was defined as the antiserum dilution which gave 0.3

absorbance units. The pre-immune control sera gave maxi- mum absorbance values of <O. 1 units under these conditions. The sera collected 13 weeks after primary immunization were used to further characterize sensitivities and specificities both in indirect and direct EIA.

Indirect Enzyme Immunoassay

Microtitre plates were coated overnight (room temperature) with oxacillin-GO, cloxacillin-GO, or dicloxacillin-GO, respectively, in appropriate concentrations (0.2-0.5 pg ml-l protein) giving about 1.0 k 0.1 absorbance units for the negative control. Free protein-binding sites of the plate were blocked for 30 min, then the plate was washed and made semi-dry. To each well, 50 pl of antibiotic standard solution and 50 p.1 of antiserum (rabbits 1,2, or 3 in a dilution of 1 + 899 to 1 + 7999 with PBS) were added and incubated for 2 h at room temperature. The plate was washed with 4 x 200 pl per well wash solution, goat anti-rabbit IgG-HRP conjugate solution was added, and incubated for 1 h at room temper- ature. The plate was washed again, and enzyme substrate. solution was added. After 15 min, the enzyme reaction was stopped, and the absorbance at 450 nm was measured. To determine the individual test specificity, standard curves for oxacillin, cloxacillin, and dicloxacillin were established in each test system, using four-fold determinations of each standard concentration. The relative cross-reactivity of each antibiotic was calculated on the basis of the concentration necessary to inhibit 50% binding of the labelled antigen in each test system. Fifteen other @-lactam antibiotics, as listed under Chemicals, Buffers and Equipment, were also tested for competitive antibody binding in concentrations up to 10 pg ml-1.

Direct Enzyme Immunoassay

The sera of the three rabbits from week 13 were precipitated with (NH4)2S04 (70% saturated) and dialysed against PBS (3 x 5 1) according to the method of Hebert et a1.16 A microtitre plate was coated with appropriate dilutions (1 + 499 to 1 + 999) of antiserum (100 pl) and incubated overnight at room temperature. Free protein-binding sites of the plates were blocked for 30 min, then the plate was washed and made semi-dry. To each well, 50 pl of antibiotic standard and 50 pl of antibiotic-HRP solution (cloxacillin-HRP, 1 pg ml- 1 , oxacil- lin- and dicloxacillin-HRP, 10 to 40 pg ml-1) was added and incubated for 2 h at room temperature. The plate was washed, and further treated as described for the indirect assay.

Analysis of Milk Samples

Standard solutions of cloxacillin or dicloxacillin were added to pasteurized milk samples (fat content approximately 3.8%). The samples were centrifuged (3000g; 15 min) and the aqueous phase was directly assayed by direct EIA using serum from rabbit 2, cloxacillin-HRP, and a cloxacillin standard curve. The EIA test protocol was essentially the same as described above, except that cloxacillin standards were prepared daily in distilled water containing 100 g of skimmed milk powder 1-1. All standards and samples were analysed in quadruplicate. All recovery experiments were performed at least 4 times on different days.

Results Antibodies reactive with the cloxacillin-HRP conjugate could be detected in all three rabbits immunized with the cloxacillin- HSA conjugate, the individual antibody titres are shown in Fig. 2. The sera obtained from rabbits 2 and 3 from week 13 through 31 were similar in aspects of antibody dilution for

Publ

ishe

d on

01

Janu

ary

1994

. Dow

nloa

ded

by U

nive

rsity

of

Win

dsor

on

27/1

0/20

14 0

4:32

:22.

View Article Online

Analyst, December 1994, Vol. 11 9 2767

EIA, whereas the sera from rabbit 1 had comparatively low antibody titres, but still sufficient for use in an EIA.

The measuring ranges of the EIA standard curves for oxacillin, cloxacillin, and dicloxacillin, established with each of the three antisera, were strongly dependent on the labelled antigen used for competition with the free drugs. The individual 50% inhibition values for oxacillin, cloxacillin and dicloxacillin, which were used to characterize test sensitivity and specificity, are summarized in Table 1. None of the 15 other penicillin antibotics listed under Chemicals, Buffers and Equipment revealed cross-reactivity at concentration levels as high as 10 pg ml-1.

The recoveries for milk samples artificially contaminated with cloxacillin or dicloxacillin, obtained in a direct EIA using

serum of rabbit 2 and cloxacillin-HRP conjugate, are sum- marized in Table 2. Both compounds could be detected at or below the MRL concentration level of 30 ng ml-1. The recoveries obtained for cloxacillin and dicloxacillin (expressed as cloxacillin equivalents) at levels of 10 ng ml-1 and 30 ng ml-1 were at 102 and 84%, respectively. No false positive results for de-fatted pasteurized milk samples were observed throughout the study when the cloxacillin standard curve was performed in an aqdeous solution of skimmed milk powder. The average relative absorbance for antibiotic free milk was 97.1 k 7.8%. Owing to the weak cross-reaction of oxacillin in the test system, the sensitivity for this compound was too low to enable detection of this compound in milk at the 30 ng ml-1 level.

100

80 = 9 60 !I?

0

- 40

20

0' I I I 5 15 25 35

Weeks after primary immunization

Fig. 2 Relative antiserum titres of the three rabbits (A, R1, B, R2, and C, R3) immunized with cloxacillin-HSA. Vertical bars indicate booster injections.

Discussion

The aim of this study was to produce antibodies with limited group specificity for the closely related isoxazolyl penicillins within the large group of (3-lactam antibiotics. Thus, both the choice of the antigen and the conjugation method for immunogen synthesis were critical factors. Cloxacillin was chosen as the antigen for two reasons. Firstly, it is the isoxazolyl penicillin that is most frequently used in the prophylaxis of bovine mastitis, and therefore was given priority within this group of antibiotics. Secondly, cloxacillin has an intermediate position between oxacillin (no chlorine substituent) and dicloxacillin (two chlorine atoms), making cross-reactivity of anti-cloxacillin antibodies with both other compounds very likely. As the carboxyl group was used as the site of conjugation, the side chain residue, i .e . , the isoxazolyl

Table 1 Concentrations of oxacillin (OC), cloxacillin (CC). and dicloxacillin (DC) required for SOYO binding in indirect (GO conjugates) and direct (HRP conjugates) EIAs, using antiserum of rabbits 1, 2, or 3. nd = Not determined

Serum rabbit 1 Serum rabbit 2 Serum rabbit 3 and standard/ and standard/ and standard/

ng ml-L ng ml-1 ng ml-1

Labelledantigen OC CC DC OC CC DC OC CC DC Indirect EIA OC-GO 268 17 526 297 7 31 364 21 126 CC-GO 362 27 486 478 21 81 1460 28 SO5 DC-GO 2020 61 134 524 39 39 3900 37 90 Direct EIA OC-HRP nd nd nd nd nd nd 278 37 119 CC-HRP 1540 76 1660 498 21 34 510 80 133 DC-HRP nd nd nd nd nd 1320 81 83

Table 2 Recovery of cloxacillin (CC) and dicloxacillin (DC) from artificially contaminated milk samples (results expressed as cloxacillin equivalents): s, standard deviation; s,, relative standard deviation; and n, number of replicates

Relative EIA Cloxacillin equivalents absorbance recovered

Antibiotic value, BIBo x 100 added/ Mean s s r Mean/ s/ s, Recovery ng ml-1 (Yo) (Yo) (YO) ngml-1 ngml-1 (Yo) (%) n 0 97.1 7.8 8.0 Negligible* 24 cc 10 50.0 8.7 17.4 10.2 3.0 29.7 102 7 cc 20 41.5 8.8 21.2 16.9 4.1 23.9 85 4 CC 30 30.1 3.8 12.6 30.7 5.0 16.4 102 13 CC 40 24.1 4.3 18.0 40.5 4.3 10.7 101 4 cc so 19.9 5.3 28.2 53.9 7.7 14.2 108 4 CC 60 16.9 3.8 22.6 66.1 8.1 12.2 110 8 DC 30 35.2 4.2 11.8 25.3 3.7 14.5 84 6 DC 60 24.2 3.6 15.2 48.0 10.4 21.7 80 4

* Absorbance values SOYO BIBo (corresponding to approximately 2 ng of cloxacillin ml-1) were considered to be out of the measuring range of the standard curve (Fig. 3).

Publ

ishe

d on

01

Janu

ary

1994

. Dow

nloa

ded

by U

nive

rsity

of

Win

dsor

on

27/1

0/20

14 0

4:32

:22.

View Article Online

2768 Analyst, December 1994, Vol. 11 9

ring system, is the most important epitope for antibody binding, thus eliminating cross-reaction with other p-lactam antibiotics.

In order to determine the effect of homologous and heterologous labelled antigens and EIA test format (direct, indirect) on test sensitivity and specificity, H R P conjugates and glucose oxidase conjugates of the three isoxazolyl penicillins were prepared. The concentration df antibiotic necessary in each test combination to give 50% binding showed wide variations from 268 to 3900 ng ml-1 (oxacillin), from 7 to 81 ng ml-1 (cloxacillin), and from 31 to 1660 ng ml-1 (dicloxacillin), respectively. Especially for the sera of rabbits 1 and 3, large differences in the sensitivities for oxacillin, cloxacillin, and dicloxacillin were observed with different competing labelled antigens. For example, oxacillin-GO as the solid-phase antigen gave nearly 10 times higher sensitivity for oxacillin standard solutions than dicloxacillin-GO. The differences obtained for serum 2 were less pronounced but still obvious. In general, maximum sensitivity for oxacillin and cloxacillin was achieved when oxacillin-GO was used, where- as the use of dicloxacillin-conjugates as the labelled antigen gave the highest sensitivity for dicloxacillin. Thus, the possibility of manipulating the test specificity through the choice of the labelled antigen was demonstrated for the isoxazolyl penicillins. Further studies are required as to whether it is possible to further increase the sensitivity towards oxacillin.

When stored frozen at - 18 “C, the HRP and G O conjugates were found to be stable for several months, upto at least one year if lyophilized. On studying the individual working concentrations however, which also determine the practical use of the labelled antigens, marked differences between GO and H R P conjugates were found. All GO conjugates could be used in dilutions corresponding to concentrations of less than 1 yg ml-1, a single batch as prepared here being sufficient to coat several thousand microtitre plates. The cloxacillin-HRP was used in a concentration of 1 pg ml-1 for EIA, which is still in a concentration range obtained for other antimicrobial compounds. 17 Conversely, the HRP conjugates of oxacillin and dicloxacillin had to be used in concentrations exceeding 10 yg ml-1 to obtain sufficient absorbance values. This greatly limited the practical use of the latter two conjugates, as high HRP concentrations also increase the risk of unspecific binding reactions owing to the excess of non-conjugated peroxidase probably present in these conjugates. These conjugates were therefore only tested with one serum to check

0.8 0 m d + 0.6 a,

s 0.4 5 2 0.2

0

80 x g 60 -

5 40 a

P

a, .- c 20 5

rf

0 0.1 1 10 100

Cloxacillidng mi-’

Fig. 3 Standard curve for cloxacillin in the direct EIA (serum rabbit 2, cloxacillin-HRP) for the detection of cloxacillin and dicloxacillin in milk. The standards (four-fold determinations) were performed in PBS containing 100 g of skimmed milk powder 1 - 1 . The absorbance value of the negative control solution (B) was 1.044 units. The 50% inhibition concentration (14 ng ml-1) is indicated by arrows.

possible changes in sensitivity. However, the 50% binding values indicated that no sensitivity improvement could be achieved with these conjugates.

As the direct EIA is, in practice, more convenient to perform when compared with the indirect assay, a combina- tion of serum from rabbit 2 and cloxacillin-HRP was chosen for an initial application study. As the analytical ‘gap’ between the sensitivity provided by the test and the sensitivity required by the MRLs did not allow dilution of milk samples to overcome matrix effects, several diluents for cloxacillin standards were tested, in order to obtain standard curves identical to those performed in de-fatted milk (data not shown). Optimal results were obtained when cloxacillin standards were performed in a solution containing 10% skimmed milk powder. A typical standard curve is shown in Fig. 3. All antibiotic-free milk samples gave negative results. The results of the recovery studies for cloxacillin and dicloxacillin showed that both compounds could easily be detected at concentrations at o r below the MRLs. Although the maximum sensitivity for cloxacillin in milk would be in the range of 3 ng ml-1, only concentrations giving relative absorbance values <50% of the negative control (approxi- mately 10 ng of cloxacillin per ml) were tested, in order to avoid ambiguous results. The test sensitivity for oxacillin was insufficient to detect this compound in milk at the MRL value of 30 ng ml-1 without a sample extraction and a concentration step. From the relative cross-reaction of oxacillin, the estimated detection limit for this compound in milk would be in the range of 100 ng ml-1. However, the test could be very useful to qualitatively detect oxacillin, cloxacillin, and diclox- acillin in meat samples, as the MRL in these matrices is comparatively high (300 ng g-1).

1 2

3

4

6 7

8

9 10

11

5

12 13

14

15

16

17

References Gedek, W . , Tieraerztl. Umsch.. 1979, 34, 4 . Commission Regulation (EEC) No. 675/92, Off. J. Eur. Commun, 1992, L 73, 8. Brynes, S. D.. and Yong, M. S . , in Euroresidue I I , eds. Haagsma, N . , Ruiter, A . , Czedik-Eysenberg, P. B . , University of Utrecht, The Netherlands, pp. 627-631. Muller. F. J . , and Jones, A . , Bull. Int. Dairy Fed., 1993,283,24. Boison, J . O . , J. Chromatogr., 1992, 624, 171. Shaik. B . and Moats, W . A . , J. Chromatogr., 1993, 643, 369. Usleber, E . , Martlbauer, E . , Schneider, E . , and Dietrich, E. , Arch. Lebensmittelhyg. 1994, 45, 32. Jackman, R. , Mitchell. S. J . , Dyer, S. D . , and Chesham. J . , Food Agric. Immunol., 1991,3, 3. AOAC Research Institute, J . AOAC Int., 1993, 76, 171A. Van Weemen, B. K . , and Schuurs, A . H. W . M.. Immuno- chemistry 1975, 12,667. Martlbauer, E . , Enzymirnmuntests fur autimikrobiell Wirksame Sfoffe, 1993, Ferdinand Enke Verlag, Stuttgart, Germany, pp. 206-222. Erlanger, B. F . , Methods Enzymol. 1980, 70, 85. Lowry, 0. H . , Rosebrough, N. J . , Farr A . L., and Randall, R . J . , J. Biol. Chem., 1951, 193, 265. Kitagawa, T., Shimozono, T. , Aikawa, T . , Yoshida, T . , and Nishimura, H . , Chem. Pharm. Bull., 1981,29, 1130. Martlbauer, E . , Gareis, M., and Terplan, G . , J. Appl. Environ. Microbiol., 1988. 54, 225. Herbert, G . A . , Pelham, P. L., and Pittman, B . , Appl. Microbiol. 1973, 25, 26. Martlbauer, E . , Meier, R . , Usleber, E. , andTerplan, G. , Food Agric. Immunol., 1992, 4, 219.

Paper 4/03397C Received June 6, 1994

Accepted August 26, 1994

Publ

ishe

d on

01

Janu

ary

1994

. Dow

nloa

ded

by U

nive

rsity

of

Win

dsor

on

27/1

0/20

14 0

4:32

:22.

View Article Online