Studies on whole cell fluorescence-based screening for epoxide hydrolases and Baeyer-Villiger monoxygenases

ARTICLE

Studies on whole cell fluorescence-based screening for epoxide hydrolases and Baeyer-Villiger monoxygenases

Beatriz Bicalho^I; Lu S. Chen^I; Johann Grognux^II; Jean-Louis Reymond^II; Anita J. Marsaioli^*, I

^IInstituto de Química, Universidade Estadual de Campinas, CP 6154, 13083-970 Campinas - SP, Brazil

^IIUniversity of Berne, Department of Chemistry and Biochemistry, Freiestrasse 3, CH-3012, Berne, Switzerland

ABSTRACT

Biocatalysis reactions were performed on microtiter plates (200 µL) aiming at the utilization of fluorogenic substrates (100 µmol L^-1) for rapid whole cell screening for epoxide hydrolases (EHs) and Baeyer-Villiger monoxygenases (BVMOs). A final protocol was achieved for EHs, with 3 new enzymatic sources being detected (Agrobacterium tumefaciens, Pichia stipitis, Trichosporom cutaneum). The fluorogenic assay for BVMO did not work as expected. However, an approach to possible variables involved (aeration; pH) provided the first detection of a BVMO activity in T. cutaneum.

Keywords: microorganisms, Baeyer-Villiger monoxygenase, epoxide hydrolase

RESUMO

Reações de biocatálise foram realizadas em microplacas (200 µL) visando a utilização de substratos fluorogênicos (100 µmol L^-1) para prospecção rápida de epóxido hidrolases (EHs) e Baeyer-Villiger monoxigenases (BVMOs) em microrganismos (células inteiras). Um protocolo final foi alcançado para EHs, com a detecção de 3 novas fontes enzimáticas (Agrobacterium tumefaciens, Pichia stipitis, Trichosporom cutaneum). O ensaio fluorogênico para BVMO não ocorreu como esperado. A abordagem de algumas variáveis envolvidas (aeração; pH) proporcionou a detecção inédita da atividade enzimática de BVM em T. cutaneum.

Introduction

A challenge that chemists are presently facing is the development of more efficient and cleaner routes to make old and new products. It is here that catalysts such as enzymes are unique making the best of raw material saving energy and replacing toxic reagents. Notwithstanding these qualities enzymes can display high stereoselectivities and that is why biotechnological processes are developing rapidly in industrial organic synthesis. It is also true that efficient biocatalysts are nowadays limited because industrial substrates are usually artificial and suitable enzymes are often unknown. Therefore, special biocatalysts must be found by screening microorganisms or any other biological source rich in enzymes and this have been a reason for several achievements in catalysis assays applied to screening enzyme activity in high throughput format with liquid chromatography-mass spectrometry (LC-MS), capillary electrophoresis, and assays based on fluorogenic or chromogenic substrates.^1-5

The presence of useful biocatalysts in microorganisms of two Brazilian Culture Collections (CCT — Coleção de Culturas Tropical, Fundação Tropical de Pesquisas André Tosello; CBMAI — Coleção Brasileira de Microrganismos de Ambiente e Indústria, Centro Pluridisciplinar de Pesquisas Químicas, Biológicas e Agrícolas, UNICAMP) have been investigated.^6-8 However, the time-consuming procedures typical of classical methodologies for screening whole cell enzymatic activity have been a critical drawback. Although high throughput screening (HTS) assays are being rapidly innovated,^2,9-11 whole cell HTS assays have been almost neglected, excepting those based on fluorogenic substrates which have been applied to metagenomic libraries (in function-driven analysis) but seldom to natural culture collections. In order to address this problem, this paper reports the application of a 96-well microplate fluorogenic assay, originally designed for purified hydrolytic enzymes,^12,13 to screen microbial whole cells for epoxide hydrolases (EHs) and Baeyer-Villiger monooxygenases (BVMOs).

Results and Discussion

The experimental approach herein described was adapted from the high throughput screening assay methodology introduced by Reymond and Klein¹² (Scheme 1) and by Furstoss and co-workers¹⁴ (Scheme 2). Epoxides, probe-substrates for EH, were known to endure the assay conditions.^4,12 The same was not true for ketone 9, therefore its stability in borate buffer pH 8.8 had to be evaluated in the presence and in the absence of the assay components (Scheme 2), namely horse liver alcohol dehydrogenase (HLADH), NAD⁺ and BSA. No time-dependent increase in fluorescence was observed for any combination of reagents within 48 h, discarding the formation of 11 and release of 7 in the absence of a specific Baeyer-Villiger monooxygenase.

In order to validate the oxidation of 11 by HLADH/NAD⁺ and subsequent b-elimination step (Scheme 2), the fluorescence of a borate buffer solution (pH 8.8) of 11 and BSA with or without HLADH/NAD⁺ was evaluated. A significant time-dependent increase in fluorescence intensity was only observed in the presence of HLADH/NAD⁺, thus indicating the release of 7 due to the oxidation of 11.

To adjust the enzymatic assay to the cell format, fixed amounts of diol 3, metaperiodate (NaIO₄), and BSA (see Scheme 1) were set to react in the presence of increasing concentrations of Rhodotorula glutinis CCT 2182. As shown in Figure 1, the time-dependent increase in fluorescence intensity due to 7 released as final product was inversely proportional to cell concentration. The presence of 0.1 mg mL^-1 cells in the solution was not detrimental to the lecture, as estimated from the control reaction. Thus, this limit was established as the maximum cell concentration to run the screening assays. The microorganisms used (Table 1) were cultivated for 3 days on appropriate agar media, at room temperature. The colonies were collected and suspended in borate buffer pH 8.8 just prior to the reactions, which were performed without cell ´ substrate pre-incubation.

Screening for EH

R. glutinis was included among the 11 microrganisms as a positive control due to its known EH activity.^6,15,16 The reactions with 1 and 2 (Scheme 1) were monitored for 10 h. The time-dependent increase in fluorescence intensity revealed three different EH biocatalysts. The EH activity revealed by T. cutaneum CCT 1903 was selective for 1 while that of A. tumefaciens CCT 6515 or P. stipitis CCT 2617 was selective for 2. On the other hand, the EH activity of R. glutinis CCT 2182 did not discriminate between 1 and 2 (Figures 2 and 3). T. cutaneum, A. tumefaciens or P. stipitis have never been mentioned before as EH sources, and we are now investigating the site and stereoselectivity of the detected biocatalysts.

Screening for BVMO

Following the same strategy described for the EH, fungus G. candidum CCT 1205 was included among the investigated microorganisms, as it was known to convert methyl-cyclohexanones into methyl-e-caprolactones.¹⁷ However, monitoring BVMO activity applying the fluorogenic HTS protocol (Scheme 2) was not as straightforward as expected. No significant time-dependent increase in fluorescence was detected, and even after 48 h of incubation only a very discrete fluorescence increment was registered for the reaction with T. cutaneum CCT 1903. Assigning these results to a putative low oxygenation level of the microtiter plate, traditional resting cell biocatalysis of ketone 9 into ester 10 or alcohol 11 (Scheme 2) by G. candidum CCT 1205 or T. cutaneum CCT 1903 was monitored by HPLC. In these reactions the release of 7 was not expected as no HLADH/NAD⁺ and BSA were added to the medium. However, in accordance with the previous results, no time-dependent significant increase in compounds 10 or 11 was detected (Table 2).

As the BVMO of G. candidum CCT 1205 was previously detected in aqueous phosphate (pH 6.5),¹⁷ the buffer influence on the BVMO activity was investigated by performing the reaction in this condition. However, the conversion of 9 into 11 by G. candidium CCT 1205 in aqueous phosphate (pH 6.5) did not even reach 3% within 48 h. On the other, the lower pH was found essential to activate the BVMO of T. cutaneum CCT 1903, as almost 30% of the starting 9 was converted into 11 within 48 h. In both cases the intermediate 10 was never detected, probably due to the action of hydrolytic enzymes such as lipases or esterases, as previously mentioned.¹⁴

As reviewed recently, BVMOs are classified according to substrate specificity,¹⁸ therefore BVMOs of G. candidum CCT 1205 and T. cutaneum CCT 1903 belong to different groups. Therefore, probe-substrates for enzyme activity must be carefully selected as "you get what you screen for". Taking this in consideration we have tested both microorganisms in whole cell biocatalysis experiments using several ketones confirming the presence of CHMO in G. candidum and BVMO for special alkanones in T. cutaneum, these results will be published elsewhere.

The oxidation of 11 by HLADH/NAD⁺ in phosphate buffer pH 6.5 did not work satisfactorily, preventing the application of the Scheme 2 to screening for BVMO activity.

Experimental

7-(2'-oxyranyl-ethoxy)-2H-1-benzopyran-2-one, (1)

A mixture of 7-hydroxy-2H-1-benzopyran-2-one (0.91 g, 5.6 mmol), NaH (60% suspension in oil; 0.61 g, 15 mmol) and 3-butenyl p-toluenosulfonate (1.14 g, 5 mmol) were stirred in DMF (20 mL) for 24 h. The mixture was then diluted with AcOEt and washed successively with H₂O and 1 mol L^-1 aqueous NaOH. Evaporation of the organic phase and flash chromatography (FC; hexane/AcOEt gradient) gave the olefinic precursor of 1, which (0.10 g, 0.46 mmol) was dissolved in CH₂Cl₂ (3 mL) containing MCPBA (0.10 g, 0.55 mmol). After 24 h at ambient temperature, the mixture was diluted with CH₂Cl₂ and washed successively with 10% aqueous Na₂SO₃ and 0.1 mol L ^-1 aqueous sodium bicarbonate. Evaporation of the organic phase and FC (hexane/EtOAc gradient) gave 1 (0.85 g, 0.37 mmol, 80%) as colorless crystals; mp 67.8 - 69.3ºC (EtOAc). lit.¹² 60 - 64 ºC. ¹H NMR (300.067 MHz, CDCl₃, d_TMS 0.00 ppm) d 7.45 (1 H, d, ³J 9.2), 7.38 (1 H, d, ³J 8.4), 6.85 (1 H, dd, ³J 8.4; 2.5), 6.83 (1 H, d, ³J 2.5), 6.26 (1 H, d, ³J 9.2), 4.19 (2 H, m), 3.17 (1 H, m), 2.86 (1 H, t, ³J 4.7), 2.60 (1 H, dd, ³J 4.7; 2.7), 2.19 ( 1 H, m), 1.95 (1 H, m); ¹³C NMR (75.45 MHz, CDCl₃, d_CDCl3 77.0 ppm) d 161.7 (C), 161.0 (C), 155.7 (C), 143.2 (CH), 128.7 (CH), 113.1 (CH), 112.7 (CH), 112.6 (C), 101.4 (CH), 65.3 (CH₂), 49.5 (CH), 47.2 (CH₂), 32.3 (CH₂); EIMS ,m/z 232 (M⁺, 86%), 162 (42), 134 (100), 89 (29), 71 (34).

A similar procedure starting with cis-3-hexenyl p-toluenosulfonate (1.66 g, 6.5 mmol) and 7-hydroxy-4-methyl-2H-1-benzopyran-2-one (1.26 g, 7.1 mmol) gave a diasteroisomeric mixture of cis/trans (1.5:1.0) 2 as a pale oil (1.26 g, 73%). ¹H NMR (499.883 MHz, CDCl₃, d_TMS 0.00 ppm), cis-isomer, d 7.50 (1 H, d, ³J 8.7), 6.87 (1 H, dd, ³J 8.7; 2.5), 6.83 (1 H, d, ³J 2.5), 6.15 (1 H, s), 4.20 (2 H, m), 3.19 (1H, ddd, ³J 7.3, 4.6, 4.6), 2.99 (1H, ddd, ³J 6.4, 6.4, 4.3), 2.15 (1 H, m), 1.96 (1 H, m), 1.09 (3H, t, ³J 7.6); trans-isomer, d 7.50 (1 H, d, ³J 8.7), 6.87 (1 H, dd, ³J 8.7; 2.5), 6.83 (1 H, d, ³J 2.5), 6.15 (1 H, s), 4.16 (2 H, m), 2.94 (1H, ddd, ³J 6.2, 4.6, 2.1), 2.79 (1H, ddd, ³J 5.5, 5.5, 2.1), 1.61 (2 H, m), 1.01 (3H, t, ³J 7.6); ¹³C NMR (75.45 MHz, CDCl₃, d_CDCl3 77.0 ppm) cis-isomer, d 161.8 (C), 161.2 (C), 155.3 (C), 152.4 (C), 125.6 (CH), 113.7 (C), 112.4 (CH), 112.0 (CH), 101.7 (CH), 65.8 (CH₂), 58.3 (CH), 54.3 (CH), 27.7 (CH₂), 21.6 (CH₂), 18.6 (CH₃), 10.4 (CH₃); trans-isomer, d 161.8 (C), 161.2 (C), 155.3 (C), 152.4 (C), 125.6 (CH), 113.7 (C), 112.4 (CH), 112.0 (CH), 101.7 (CH), 65.2 (CH₂), 60.1 (CH), 55.4 (CH), 31.8 (CH₂), 24.9 (CH₂), 18.6 (CH₃), 9.8 (CH₃); cis-isomer, EIMS: m/z 274 (M⁺, 26%), 256 (7), 188 (30), 176 (58), 148 (100); trans-isomer, EIMS: m/z 274 (M⁺, 100%), 256 (12), 188 (18), 176 (44), 148 (46).

7-(4'-oxo-pentyl)-2H-1-benzopyran-2-one, (9)

A mixture of 7-hydroxy-2H-1-benzopyran-2-one (0.57 g, 3.5 mmol), NaH (60% suspension in oil; 0.4 g, 10 mmol) and 4-pentenyl p-toluenosulfonate (0.8.4 g, 3.5 mmol) were stirred in DMF (15 mL) for 24 h. The mixture was then diluted with EtOAc and washed successively with H₂O and 1 mol L^-1 aqueous NaOH. Evaporation of the organic phase and FC (hexane/AcOEt gradient) gave the olefinic precursor of 9, which (0.16 g, 0.69 mmol) was dissolved in a mixture of PdCl₂ (12.1 mg, 0.07 mmol) and CuCl (68.3 mg, 0.69 mmol) in DMF-H₂O (5 mL, 7:1) under O₂ atmosphere. After 24 h at ambient temperature the mixture was diluted with 5% aqueous HCl and washed with EtOAc. Evaporation of the organic phase and FC (hexane/AcOEt gradient) gave 9 (0.11 g, 64%) as colorless crystals; mp 74.4 - 75.8 ºC (EtOAc). ¹H NMR (300.067 MHz, CDCl₃, d_TMS 0.00 ppm) d 7.62 (d, 1 H, ³J 9.5), 7.35 (d, 1 H, ³J 8.2), 6.80 (dd, 1 H, ³J 8.2; 2.5), 6.76 (d, 1 H, ³J 1.8), 6.22 (d, 1 H, ³J 9.5), 4.02 (t, 2 H, ³J 6.0), 2.66 (t, 2 H, ³J 7.1), 2.17 (s, 3 H), 2.08 (2 H, qt, ³J 6.6); ¹³C NMR (75.45 MHz, CDCl₃, d_CDCl3 77.0 ppm) d 207.5 (C), 161.8 (C), 160.9 (C), 155.5 (C), 143.2 (CH), 128.6 (CH), 112.9 (CH), 112.5 (CH), 112.4 (C), 101.3 (CH), 67.4 (CH₂), 39.6 (CH₂), 30.1 (CH₂), 23.0 (CH₃); EIMS: m/z 246 (M⁺, 5%), 162 (7), 134 (31), 85 (100), 43 (75).

7-(3'-acetoxy-propyl)-2H-1-benzopyran-2-one, (10)

Acetyl chloride (0.69 g, 8.5 mmol) was added to 3-propanol p-toluenosulfonate (1.15 g, 5 mmol) and pyridine (1.5 mL, 17.5 mmol) in CH₂Cl₂ (6 mL). After 30 min the mixture was diluted with EtOAc (50 mL) and washed successively with 0.1 mol L^-1 aqueous NaHCO₃, 10% aqueous CuSO₄ and H₂O. Evaporation of the organic phase and FC (hexane/AcOEt gradient) gave the p-toluenosulfonil precursor of 10, which (1.06 g, 3.9 mmol) was stirred with 7-hydroxy-2H-1-benzopyran-2-one (0.66 g, 4.1 mmol) and NaH (60% suspension in oil; 0.62 g, 12 mmol) in DMF (35 mL) for 24 h. The mixture was then diluted with EtOAc and washed successively with H₂O and 1 mol L^-1 aqueous NaOH. Evaporation of the organic phase and FC (hexane/EtOAc gradient) gave 10 (0.26 g, 26%) as pale crystals; mp 67.9 _ 68.9 ºC (EtOAc). ¹H NMR (300.067 MHz, CDCl₃, d_TMS 0.00 ppm) d 7.64 (d, 1 H, ³J 9.3), 7.37 (d, 1 H, ³J 8.4), 6.84 (dd, 1 H, ³J 8.4; 2.4), 6.80 (d, 1 H, ³J 2.4), 6.25 (d, 1 H, ³J 9.3), 4.27 (t, 2 H, ³J 6.2), 4,11 (t, 2 H, ³J 6.2), 2.16 (qt, 2 H, ³J 6.2), 2.07 (s, 3 H); ¹³C NMR (75.45 MHz, CDCl₃, d_CDCl3 77.0 ppm) d 170.7 (C), 161.7 (C), 160.9 (C), 155.6 (C), 143.2 (CH), 128.6 (CH), 113.0 (CH), 112.7 (CH), 112.5 (C), 101.2 (CH), 64.9 (CH₂), 60.9 (CH₂), 28.4 (CH₂), 21.0 (CH₃);EIMS: m/z 262 (M⁺, 100), 219 (7), 162 (29), 134 (72), 101 (83), 43 (89). The synthesis of compounds 3 and 11 were described elsewhere.¹³

Microorganisms

Yeasts and bacteria were cultured for 3 days, at room temperature, in 1.5 mL Eppendorf flasks respectively containing yeast-malt-agar medium (3 g L^-1 yeast extract Merck, 3 g L^-1 malt extract Merck, 5 g L^-1 peptone Difco, 10 g L^-1 glucose, 20 g L^-1 agar Merck, H₂O) and Müller Hinton-agar medium (20 g L^-1 Müller Hinton Merck, 20 g L^-1 agar Merck, H₂O). Just prior to the tests the grown colonies were weighed into 1.5 mL Eppendorf flasks and suspended in aqueous buffer.

Fluorescence measurements

All buffers and solutions were prepared using MilliQ-deionized H₂O. Microbial cells were diluted from 0.2 mg mL^-1 suspentions in 20 mol L^-1 borate buffer (pH 8.8). Substrates were diluted from 20 mmol L^-1 stock solutions in 50% aqueous MeCN. BSA was diluted from a 40 mg.mL^-1 stock solution in 20 mmol L^-1 borate (pH 8.8). For the EH assay, NaIO₄ was diluted from a 4.3 mg mL^-1 aqueous solution. For BVMO assay, HLADH was diluted from a 10.6 mg mL^-1 aqueous stock solution and NAD⁺ was diluted from a 13 mg mL^-1 aqueous stock solution. For screening EH activity, reactions were initiated by addition of NaIO₄ to a solution containing substrate and BSA. For screening BVMO activity, reactions were initiated by addition of cells to a solution containing substrate, HLADH, NAD⁺ and BSA. The 200 µL assays were followed in individual wells of flat-bottom polypropylene 96-well microtiter plates (Costar) with a Cytofluor-II fluorescence plate reader (Perseptive Biosystems, filters l_ex 360 ± 20 nm, l_em 460 ± 20 nm).

Biocatalysis reactions

Microbial cells weighed into 100 mL conic flasks (Costar) were suspended at 2 mg mL^-1 in aqueous buffer (20 mmol L^-1 borate pH 8.8 or 20 mmo L^-1 phosphate pH 6.5, 25 mL). Then, a solution of ketone 3 (50 µL, 100 mg mL^-1 in MeCN) was added to provide 0.2 mg mL^-1 substrate solutions and the resulting mixtures were agitated on a plate stirrer at room temperature. To follow the reactions, 20 µL aliquots were taken and analyzed by HPLC on a Chromolith Speed Rod column (RP-18e, 5.0 × 10 cm, Merck, flow 3.0 mL min^-1, 100% A to 100%B in 10 min, A = 100% H₂O + 0,1% TFA, B = 60% MeCN + 40% H₂O + 0,1% TFA): t_R 4.1 (7), 5.1 (11), 6.3 (9), 6.8 (10) min. The integral of the peaks recorded by UV at 320 nm was used to calculate the conversion of released products.

Conclusions

An enzymatic HTS fluorogenic assay was smoothly adapted to a HTS protocol for EH of whole cells, and its application revealed 3 new microbial sources of EH (A. tumefaciens, P. stipitis, and T. cutaneum). Considering the practical application of EH in organic synthesis, the utilization of the assay presented here for gathering selective biocatalysts is strongly recommended. On the other hand, the HTS fluorogenic assay based on Furstoss's procedure,¹⁴ used for present evaluation of monooxigenase, proved to be inadequate for cyclohexanone monooxigenase. Thus, a series of modifications are being evaluated. Notwithstanding this shortcoming the experiments revealed 2 sources of BVMO biocatalysts (G. candidum and T. cutaneum).¹⁹

Acknowledgements

This work was supported by Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) and Conselho Nacional de Pesquisa e Desenvolvimento (CNPq).

References

1. Niessen, W.M.A.; J. Chromatogr. A 2003, 1000, 413.

2. Leroy, E.; Bensel, N.; Reymond, J.-L.; Bioorg. Med. Chem. 2003, 13, 2105.

3. Dean, K.E.S.; Klein, G.; Renaudet, O.; Reymond, J.-L.; Bioorg. Med. Chem. 2003, 13, 1653.

4. Wahler, D.; Badalassi, F.; Crotti, P.; Reymond, J.-L.; Angew. Chem. Int. Ed. 2001, 40, 4457.

5. Zocher, F.; Enzelberger, M.M.; Bornscheuer, U.T.; Hauer, B.; Schmid, R.D.; Anal. Chim. Acta 1999, 391, 345.

6. Cagnon, R.; Marsaioli, A.J.; Riatto, V.B.; Pilli, R.A.; Chemosphere 1999, 38, 2243.

7. Porto, A.L.M.; Cagnon, R.; Marsaioli, A.J.; Eguchi, S.Y.; Chemosphere 1999, 38, 2237.

8. Marsaioli, A.J.; de Conti, R.M.; Porto, A.L.M.; J. Mol. Cat. B. Enzym. 2001, 11, 233.

9. Ravot, G.; Wahler, D.; Favre-Bulle, O.; Cilia, V.; Lefevre, F.; Adv. Synth. Catal. 2003, 345, 691.

10. Mateo, C.; Archelas, A.; Furstoss, R.; Anal. Biochem. 2003, 314, 135.

11. Reetz, M.T.; Angew. Chem. Int. Ed. 2002, 41, 1335.

12. Klein, G.; Reymond, J.-L.; Helv. Chim. Acta. 1999, 82, 400.

13. Badalassi, F.; Wahler, D.; Klein, G.; Crotti, P.; Reymond, J.-L.; Angew. Chem. Int. Ed. 2000, 39, 4067.

14. Sleegers, A.; Simpson, H. D.; Alphand, V.; Furstoss, R.; The 5^th International Symposium on Biocatalysis and Biotransformation, Biotrans, 2001, Darmstadt , Germany.

15. Weijers, C.A.G.; Botes, A.L.; Van Dyk, M.S.; De Bont, J.A.M.; Tetrahedron: Asymmetry 1998, 9, 467.

16. Kronenburg, N.A.E.; Mutter, M.; Visser, H.; De Bont, J.A.M.; Weijers, C.A.G.M.; Biotechnol. Lett. 1999, 21, 519.

17. Porto, A.L.M.; PhD. Thesis, Universidade Estadual de Campinas, Brazil, 2002.

18. Kamerbeek, N.M.; Janssen, D.B.; van Berkel, W.J.H.; Fraaije, M.W.; Adv. Synth. Catal. 2003, 345, 667.

19. After submission of the present work we became aware of Furstoss's publication on the optimized method: Gutiérrez, M.C.; Sleegers, A.; Simpson, H. D.; Alphand, V.; Furstoss; Org. Biomol. Chem. 2003, 1, 3500 and the BVMO in G.candidum: Carballeira, J.D.; Alvarez, E.; Sinisterra. J.V.; J. Mol. Catalysis B, Enzymatic. 2004, 28, 25.

Received: September 22, 2003

Published on the web: November 12, 2004

Publication Dates

• Publication in this collection
  24 Jan 2005
• Date of issue
  Dec 2004

History

• Accepted
  12 Nov 2004
• Received
  22 Sept 2003