|Home | About | Journals | Submit | Contact Us | Français|
The red carotenoid astaxanthin possesses higher antioxidant activity than other carotenoids and has great commercial potential for use in the aquaculture, pharmaceutical, and food industries. In this study, we produced astaxanthin in the budding yeast Saccharomyces cerevisiae by introducing the genes involved in astaxanthin biosynthesis of carotenogenic microorganisms. In particular, expression of genes of the red yeast Xanthophyllomyces dendrorhous encoding phytoene desaturase (crtI product) and bifunctional phytoene synthase/lycopene cyclase (crtYB product) resulted in the accumulation of a small amount of β-carotene in S. cerevisiae. Overexpression of geranylgeranyl pyrophosphate (GGPP) synthase from S. cerevisiae (the BTS1 gene product) increased the intracellular β-carotene levels due to the accelerated conversion of farnesyl pyrophosphate to GGPP. Introduction of the X. dendrorhous crtS gene, encoding astaxanthin synthase, assumed to be the cytochrome P450 enzyme, did not lead to astaxanthin production. However, coexpression of CrtS with X. dendrorhous CrtR, a cytochrome P450 reductase, resulted in the accumulation of a small amount of astaxanthin. In addition, the β-carotene-producing yeast cells transformed by the bacterial genes crtW and crtZ, encoding β-carotene ketolase and hydroxylase, respectively, also accumulated astaxanthin and its intermediates, echinenone, canthaxanthin, and zeaxanthin. Interestingly, we found that these ketocarotenoids conferred oxidative stress tolerance on S. cerevisiae cells. This metabolic engineering has potential for overproduction of astaxanthin and breeding of novel oxidative stress-tolerant yeast strains.
The red carotenoid astaxanthin (3,3′-dihydroxy-β,β-carotene-4,4′-dione) is widely distributed in fish, birds, and crustaceans and has mainly been used as a pigmentation source for salmonid flesh in the aquaculture industry (5). Because astaxanthin shows higher antioxidant activity than β-carotene and α-tocopherol (14, 18, 23), it has great commercial potential in the pharmaceutical and food industries (12, 24). Although astaxanthin is now supplied mainly by chemical synthesis, it can also be produced by algal culture or yeast fermentation (6). At present, the green alga Haematococcus pluvialis and the basidiomycetous yeast Xanthophyllomyces dendrorhous (the telemorphic state for Phaffia rhodozyma) are used for as natural producers of astaxanthin. However, these biological production methods remain problematic due to the requirement of strong light in H. pluvialis and low productivity in X. dendrorhous. Additional applications would be feasible if the cost of astaxanthin production could be reduced to the level of the chemical synthesis method (11), and many studies have been attempted to improve astaxanthin production in these organisms (7, 26, 28).
The genes involved in astaxanthin biosynthesis have been cloned in some organisms. The marine bacterium Paracoccus sp. (formerly Agrobacterium aurantiacum) is a known astaxanthin producer (19). In this bacterium, a key step in astaxanthin synthesis involves conversion of geranylgeranyl pyrophosphate (GGPP) to phytoene by phytoene synthase encoded by crtB (Fig. (Fig.1).1). Phytoene is converted to lycopene by phytoene desaturase encoded by crtI, followed by β-carotene production through a ring-closing reaction at both ends of the lycopene by lycopene cyclase encoded by crtY. Finally, β-carotene is converted to astaxanthin via four steps in which two keto and hydroxy groups are added to each ring by β-carotene ketolase encoded by crtW and β-carotene hydroxylase encoded by crtZ, respectively. In contrast, a single X. dendrorhous gene, crtYB, encodes a bifunctional enzyme phytoene synthase/lycopene cyclase (30), and crtS encodes astaxanthin synthase, catalyzing ketolation and hydroxylation of β-carotene (22). These findings suggest that only three genes are required for astaxanthin production from GGPP in X. dendrorhous. On the basis of the predicted protein sequence, crtS is presumed to encode a cytochrome P450 protein (3, 22), and crtS-disrupted cells of X. dendrorhous do not synthesize astaxanthin. However, the CrtS protein has not been characterized, and its heterologous expression has been unsuccessful (17). Recent investigations also strongly support involvement of a cytochrome P450 reductase gene, crtR, in astaxanthin synthesis in X. dendrorhous (2).
Verwaal et al. (31) reported that the budding yeast Saccharomyces cerevisiae accumulated β-carotene as a result of the introduction of the carotenogenic genes of X. dendrorhous. In that report, X. dendrorhous CrtYB and CrtI were overexpressed in S. cerevisiae, and overproduction of β-carotene was successfully achieved by the subsequent metabolic engineering. However, Verwaal et al. did not introduce crtS, and there was no description of the astaxanthin production. Therefore, astaxanthin production in S. cerevisiae by means of heterologous expression from other carotenoid-producing organisms has not yet been reported. One of our objectives in this study is to express the CrtS protein functionally in S. cerevisiae cells.
It has been suggested that free radicals and reactive oxygen species (ROS) are quenched by endogenous ketocarotenoids, such as astaxanthin. Jayaraj and Punja reported that the transgenic carrot plant expressing the algal β-carotene ketolase gene accumulated various ketocarotenoids, including astaxanthin, and exhibited remarkable resistance to UV-B irradiation and oxidative stress in comparison with the wild-type carrot (13). Further, Escherichia coli transformant cells were reported to acquire tolerance to oxidative stress by the introduction of carotenoid synthetic genes (27). During fermentation processes, yeast cells are exposed to various stresses, including high concentrations of ethanol, high temperature, and high osmolarity, which generate ROS (8-10, 21). Oxidative stress tolerance of yeast cells might improve the fermentation performance or simplify the fermentation process for the production of alcoholic beverages or bioethanol. In this study, we constructed S. cerevisiae strains that synthesize astaxanthin by expressing X. dendrorhous and/or bacterial enzymes and observed oxidative stress tolerance in yeast cells accumulating ketocarotenoids.
X. dendrorhous strains Y989 (wild type) and Y2238-10IL (astaxanthin-overproducing mutant) were gifts from Ajinomoto Co. (Tokyo, Japan). X. dendrorhous strain ATCC 24202 was from the American Type Culture Collection (Rockville, MD). All X. dendrorhous strains were cultured at 18°C with rotary shaking (150 rpm) in a YMG medium consisting of 2% glucose, 0.3% Bacto yeast extract, 0.3% Bacto malt extract, and 0.5% Bacto peptone (Difco Laboratories, Detroit, MI).
S. cerevisiae strain INVSc1 (MATa his3D1 leu2 trp1-289 ura3-52 MATα his3D1 leu2 trp1-289 ura3-52) (Invitrogen, San Diego, CA) was used as a host. A nutrient medium, YPDA (2% glucose, 1% yeast extract, 2% peptone, 0.003% adenine hemisulfate, pH 6.5) was used for the wild-type strain. The synthetic complete (SC) medium (0.67% Bacto yeast nitrogen base without amino acids [Difco Laboratories], 2% glucose, 0.2% dropout mixture, pH 5.8) lacking an appropriate nutrient(s) was used for the screening or maintenance of transformants (1).
E. coli strain DH5α [F− ϕ80dlacZΔM15 Δ(lacZYA-argF)U169 deoR recA1 endA1 hsdR17(rk− mk+) phoA supE44 λ− thi-1 gyrA96 relA1] was used as a host for plasmid construction. Luria-Bertani (LB) complete medium (0.5% Bacto yeast extract, 1% Bacto tryptone [Difco Laboratories], 1% NaCl, pH 7.0) was used, and 50 μg per ml ampicillin was added to the LB medium if necessary.
Paracoccus sp. (NBRC 101723) was purchased from the Biological Resource Center in the National Institute of Technology and Evaluation (NBRC; Chiba, Japan) and cultured in LB medium at 20°C.
Growth was monitored by measuring optical density at 600 nm for S. cerevisiae and 660 nm for X. dendrorhous and E. coli and/or direct cell counting using a microscope with a Thoma hemacytometer.
The following 2μ-based high-copy-number plasmids were used for the construction of expression vectors in this study: pTV3, pUV2, and pHV1, containing the S. cerevisiae TRP1, URA3, and HIS3 selection markers, respectively (25), and pAD4 (supplied by J. Nikawa [Kyushu Institute of Technology, Fukuoka, Japan]) (20), which contains S. cerevisiae LEU2 as a selection marker and the ADH1 promoter/terminator. To eliminate one of the BamHI sites of pHV1, pTV3 and pHV1 were digested with PstI and a small fragment derived from pHV1 and a large fragment from pTV3 were ligated and named pHV3. All heterologous genes used in this study were inserted into pAD4 once, and each gene was excised with the ADH1 promoter and terminator and used as an expression cassette. By use of the expression cassettes, various series of expression vectors for heterologous genes were constructed. Plasmid pAD-EGFP for the expression of the green fluorescent protein (GFP) fusion protein was constructed from pAD4 and the DNA fragment of EGFP amplified from pEGFP-N1 (Clontech, Mountain View, CA). Each carotenogenic gene without the stop codon was introduced upstream of EGFP in pAD-EGFP. The plasmids constructed in this study are summarized in Table Table11.
To overexpress the enzymes involved in astaxanthin synthesis, the following DNA fragments were amplified by KOD Plus DNA polymerase (Toyobo, Osaka, Japan) and the corresponding primer sets, listed in Table Table2,2, from genomic DNA or from single-stranded cDNA: crtI, crtYB, and crtS from the X. dendrorhous ATCC 24202 cDNA; BTS1 and NCP1 from the S. cerevisiae INVSc1 genome; crtW from the Paracoccus sp. genome; and crtZ from plasmid pUCmod-crtZ (gift from C. Schmidt-Dannert, University of Minnesota, St. Paul) (16) carrying crtZ of Pantoea ananatis. The genomic DNA of S. cerevisiae and Paracoccus sp. was extracted by using Dr. GenTLE (from Yeast) (Takara Bio, Ohtsu, Japan) and Isoplant (Nippon Gene, Tokyo, Japan), respectively. The mRNA of X. dendrorhous was prepared using an RNeasy minikit (Qiagen, Valencia, CA), and the cDNA was synthesized using a PrimeScript II first-strand cDNA synthesis kit (Takara Bio) and KOD Plus DNA polymerase (Toyobo). The resultant fragments were inserted into pAD4. The nucleotide sequences of the cloned genes were confirmed by DNA sequencing.
The cells were inoculated into a 100-ml Erlenmeyer flask containing 20 ml medium and cultured at 20°C in a rotary shaker (250 rpm). Five-day-cultured cells were collected and washed twice with deionized water. The cell pellet was lyophilized in a Lyph-Lock 12 vacuum freeze drier (Labconco, Kansas City, MO). Carotenoids were extracted from the dried cells with acetone and used for analysis by reverse-phase high-performance liquid chromatography (HPLC) as described previously (28). The authentic carotenoid standards and their retention times were as follows: for astaxanthin, 8.6 min; for zeaxanthin, 10.4 min; for canthaxanthin, 12.9 min; for echinenone, 18.9 min; and for β-carotene, 25.0 min.
S. cerevisiae transformant cells cultured at 20°C for 1 day were collected by centrifugation in a 2-ml round-bottomed plastic tube and washed with 10% trichloroacetic acid (TCA). The cells were resuspended by 200 μl of 10% TCA and disrupted with 500 μl of 0.5-mm glass beads in a MultiBead Shocker (Yasui Kikai, Osaka, Japan) with eight cycles of running at 2,500 rpm for 30 s and pausing for 30 s. After 200 μl of 10% TCA was added to the tube, several holes were made by a needle at the bottom of the tube, and the pellets were separated from the glass beads by centrifugation. The resultant pellets were resuspended in 100 μl of sample buffer (100 mM Tris-HCl [pH 6.8], 4% sodium dodecyl sulfate [SDS], 9% glycerol, 0.02% bromophenol blue, 1.4 M 2-mercaptoethanol), and 100 μl of 1 M Tris-HCl (pH 8.0) was added to the mixture for neutralization. After incubation for 15 min at 60°C, the supernatant was obtained by centrifugation and the protein concentration was measured by using a Bio-Rad protein assay kit (Hercules, CA). Approximately 100 μg of total proteins was loaded onto a 10.5% SDS-polyacrylamide gel (90 by 70 by 1 mm) and electrophoresed with a constant voltage at 120 V (15). Following electrophoresis, proteins were transferred to a nitrocellulose membrane (Hybond-C; GE Healthcare, Piscataway, NJ) in blotting buffer (48 mM Tris, 385 mM glycine, 0.1% SDS, 20% methanol) (115 V, 400 mA, 1 h). The membrane was soaked in blocking buffer (25 mM Tris, 137 mM NaCl, 2.7 mM KCl, 0.3% Tween 20, 3% skim milk, pH 8.0) for 1 h. Anti-GFP (Roche Diagnostics, Basel, Switzerland) and antiporin (monoclonal antibody for yeast cell biology A-6449; Invitrogen) antibodies were used as primary antibodies, and horseradish peroxidase-fused anti-mouse immunoglobulin G (ECL Plus Western blotting reagent pack; GE Healthcare) was used as a secondary antibody. Detection was performed using the ECL Plus Western blotting detection system (GE Healthcare).
Yeast cells cultured at 20°C for 1 day were collected by centrifugation and washed with 10 mM phosphate buffer (pH 7.5). The cells were resuspended in a 0.5 μM solution of ER-Tracker blue-white DPX dye (Invitrogen) and incubated at 20°C for 30 min. After being washed, the cells were resuspended in the buffer and observed with a fluorescence microscope (Axiovert 200M; Carl Zeiss, Oberkochen, Germany).
Biosynthesis of carotenoids, including astaxanthin, is initiated by the condensation of two GGPP molecules in the carotenogenic organisms, while S. cerevisiae cells do not possess the pathway downstream of GGPP (Fig. (Fig.1).1). To express the enzymes involved in astaxanthin biosynthesis of X. dendrorhous in S. cerevisiae, the expression plasmids pTV-crtI, pUV-crtYB, and pAD-crtS, carrying crtI, crtYB, and crtS from X. dendrorhous, respectively, were introduced into S. cerevisiae INVSc1 cells in various combinations. Although there was no significant difference in colony colors among the strains (Fig. (Fig.2a,2a, sections 1 to 3), HPLC analysis detected small amounts of β-carotene in S. cerevisiae cells possessing pTV-crtI and pUV-crtYB (approximately 18 μg per g [dry weight] of cells) (Fig. (Fig.2b).2b). The transformant cells possessing pTV-crtI, pUV-crtYB, and pAD-crtS also produced β-carotene but not astaxanthin.
If a low level of β-carotene were the rate-limiting step for astaxanthin synthesis in S. cerevisiae, increased GGPP synthase activity might accelerate conversion of farnesyl pyrophosphate to GGPP, a precursor of carotenoid biosynthesis (Fig. (Fig.1).1). Toward this end, we constructed high-copy-number plasmid, pHV-BTS1, for S. cerevisiae BTS1, and strain INVSc1 harboring pTV-crtI and pUV-crtYB was transformed with pHV-BTS1. As predicted, yeast cells overexpressing GGPP synthase showed remarkable yellow color and a prominent 22-fold increase in β-carotene content (approximately 390 μg per g [dry weight] of cells), probably due to an increase in enzyme activity (Fig. (Fig.2).2). However, the introduction of crtS to yeast cells that overproduce β-carotene did not cause astaxanthin production.
To confirm the expression of the gene products, each gene involved in astaxanthin synthesis was fused with the enhanced GFP (EGFP) gene and expressed under the control of the ADH1 promoter in S. cerevisiae. By immunological detection using anti-GFP antibody (Fig. (Fig.3),3), we observed the bands corresponding to each gene product with the expected molecular mass, suggesting that the gene expression had occurred successfully in S. cerevisiae cells.
The above-mentioned results suggested that the crtS gene product did not function as astaxanthin synthase in S. cerevisiae. It can be deduced from the primary structure that CrtS is likely the cytochrome P450 enzyme. In general, the P450 oxidoreductase system consists of two or three components, and the catalytic P450 molecule requires P450 reductase to transfer electrons from NADPH (29). Recently, the P450 reductase gene (crtR) was cloned from X. dendrorhous and shown to be essential for astaxanthin biosynthesis (2). Accordingly, we examined coexpression of CrtR and CrtS in S. cerevisiae. Plasmid pAD-crtR was introduced into S. cerevisiae cells harboring pHV-BTS1-crtS, which contains the expression cassette of BTS1 and crtS, pTV-crtI, and pUV-crtYB. Interestingly, HPLC analysis showed that new peaks appeared in the cell extract of the transformant and that one of these peaks, with a retention time of 8.6 min, corresponded to that of authentic astaxanthin, judging from the retention time and the absorbance spectrum (Fig. (Fig.4).4). This result suggests that the enzymatically active CrtS proteins with the aid of CrtR would result in astaxanthin production.
In carotenogenic bacteria, two genes, crtW and crtZ, encoding β-carotene ketolase and hydroxylase, respectively, are involved in the conversion of β-carotene to astaxanthin. To confirm whether the overexpression of the crtW and crtZ gene products from Paracoccus sp. and Pantoea ananatis, respectively, contributes to astaxanthin production in S. cerevisiae, each of the plasmids pHV-crtW, pHV-crtZ, and pHV-crtW-crtZ was introduced into the β-carotene-producing cells. The transformed cells harboring pHV-crtW accumulated echinenone (β,β-carotene-4-one) and canthaxanthin (β,β-carotene-4,4′-dione), while those possessing pHV-crtZ accumulated zeaxanthin (β,β-carotene-3,3′-diol) (Fig. (Fig.5).5). When pHV-crtW-crtZ carrying crtW and crtZ was introduced, astaxanthin accumulated in the cell extracts in addition to the ketocarotenoids described above (Fig. (Fig.5).5). It is noteworthy that the astaxanthin level in the transformed cells was much higher than that observed in the cells expressing X. dendrorhous CrtS and CrtR. These results indicate that the bacterial carotenogenic genes were functionally active in S. cerevisiae cells to produce astaxanthin.
As described above, we successfully constructed the S. cerevisiae strain that produces carotenoids, including astaxanthin, by the use of yeast and bacterial carotenogenic genes. It is thought that endogenous ketocarotenoids, including astaxanthin, scavenge free radicals and ROS (14, 18, 23). For this reason, we examined the oxidative stress tolerance of the carotenoid-accumulating S. cerevisiae cells. As shown in Fig. Fig.6,6, yeast cells that accumulate zeaxanthin (harboring pTV-crtI, pUV-crtYB, pAD-BTS1, and pHV-crtZ), echinenone and canthaxanthin (harboring pTV-crtI, pUV-crtYB, pAD-BTS1, and pHV-crtW), and astaxanthin (harboring pTV-crtI, pUV-crtYB, pAD-BTS1, and pHV-crtW-crtZ) showed greater tolerance to H2O2 than noncarotenogenic cells (carrying the vectors only). In contrast, β-carotene accumulation did not confer oxidative stress tolerance to yeast transformants harboring pTV-crtI, pUV-crtYB, pAD-BTS1, and pHV3. These results showed that intracellular accumulation of ketocarotenoids is effective for yeast protection from H2O2.
In this study, we engineered S. cerevisiae strains to produce astaxanthin by introducing X. dendrorhous and/or the bacterial genes involved in astaxanthin biosynthesis. First, CrtYB (phytoene synthase/lycopene cyclase) and CrtI (lycopene synthase) derived from X. dendrorhous were overexpressed in S. cerevisiae, but the transformant accumulated only a small amount of β-carotene (Fig. (Fig.2).2). We attempted to increase the expression of BTS1 in S. cerevisiae under the control of a strong and constitutive ADH1 promoter, leading to a 22-fold increase in β-carotene production.
It is noteworthy that the introduction of crtR, encoding the cytochrome P450 reductase, is crucial for the functional expression of CrtS and astaxanthin production in S. cerevisiae. We demonstrate not only astaxanthin production in S. cerevisiae but also heterologous expression of functional CrtS. Furthermore, our results strongly suggest that the CrtS protein can catalyze both keto and hydroxyl reactions of β-carotene in S. cerevisiae. It is known that most of the human cytochrome P450 enzymes function in S. cerevisiae cells (29), suggesting that the heterologous P450 protein can cooperate with the endogenous P450 reductase Ncp1, which is the only P450 reductase in S. cerevisiae. However, overexpression of Ncp1 under the control of the ADH1 promoter was not effective for the functional expression of CrtS (data not shown). These results suggest that X. dendrorhous CrtS is a unique cytochrome P450 protein that has high specificity for its own P450 reductase. Álvarez et al. reported that the amounts of β-cryptoxanthin (3-hydroxy-β,β-carotene) and zeaxanthin were increased by the introduction of crtS in the zygomycete Mucor circinelloides, but in this case, neither canthaxanthin nor astaxanthin was accumulated (3), indicating that CrtS catalyzed only the hydroxylation reaction of β-carotene in this fungus. Possibly, the catalytic reaction of CrtS might be enhanced by combining it with P450 reductase molecules of various origins. In this connection, S. cerevisiae accumulated approximately 3 μg astaxanthin per g (dry weight) of cells when CrtS and CrtR were coexpressed, whereas in X. dendrorhous, a significantly higher level of production is observed (120 μg per g [dry weight] of cells).
In general, the cytochrome P450 and P450 reductases are colocalized in the endoplasmic reticulum (ER) to facilitate electron transfer, and in some cases, these protein molecules exist as a fusion protein (29). Therefore, we supposed that cellular localization of CrtS and CrtR also influences astaxanthin biosynthesis in S. cerevisiae. Fluorescence microscopy of EGFP-fused proteins revealed that CrtR was localized in the ER membrane, although the CrtS proteins were present not only in the ER membrane but also in the cytosol (data not shown). More localization of CrtS to the ER membrane may enhance the accumulation of astaxanthin in S. cerevisiae cells.
The bacterial β-carotene ketolase (CrtW) and hydroxylase (CrtZ), which do not require reductase activity like CrtR, were successfully expressed in S. cerevisiae cells, and the transformant accumulated more astaxanthin (29 μg per g [dry weight] of cells) relative to the transformant coexpressing CrtS and CrtR. These yields of astaxanthin production in S. cerevisiae are substantially lower than that of X. dendrorhous, and further genetic manipulation and/or culture optimization would undoubtedly increase astaxanthin production. It should be noted that S. cerevisiae cells harboring pTV-crtI, pUV-crtYB, and pAD-BTS1 accumulated more β-carotene (390 μg per g [dry weight] of cells) than the total amounts of astaxanthin and β-carotene in the wild-type cells of X. dendrorhous (270 μg per g [dry weight] of cells). Therefore, improvement of the rate of conversion from β-carotene to astaxanthin may be an effective strategy for increasing productivity of astaxanthin in S. cerevisiae.
It is known that carotenoids that scavenge ROS and that ketocarotenoids exhibit more antioxidant activity than β-carotene (14, 18, 23). Jayaraj and Punja (13) reported that transgenic carrots producing ketocarotenoids showed a remarkable tolerance to oxidative stresses. In this study, ketocarotenoid-producing S. cerevisiae cells clearly showed tolerance to hydrogen peroxide. The metabolic engineering involving the introduction of carotenogenic genes may be useful for the construction of yeast strains with not only astaxanthin accumulation but also oxidative stress tolerance. Yeast cells that accumulate only β-carotene were more sensitive to oxidative stress than were noncarotenogenic cells (Fig. (Fig.6).6). Although the colony-forming rates of the carotenogenic transformants seemed to be a little slower than that of noncarotenogenic cells, all of the carotenogenic transformants showed the same rate of growth, as shown in Fig. Fig.6.6. It was suggested that the higher level of stress tolerance of the carotenogenic cells is not due to their slower growth but is due to carotenoid accumulation in the cells. The antioxidative activity of β-carotene in the transformant may be insufficient to compensate for the deficiency of ergosterol, which is also synthesized from farnesyl pyrophosphate (Fig. (Fig.11).
In terms of biotechnological applications, we found that astaxanthin and its intermediates confer oxidative stress tolerance on yeast cells. As described above, yeast cells are exposed to various stresses during fermentation processes. Baker's yeasts are widely distributed as dried yeast and are required for the second fermentation after freezing in frozen-dough baking, indicating that yeast cells suffer from desiccation and freezing stress before fermentation. It is known that the desiccation and freezing stresses cause the intracellular ROS generation (8, 10). Hence, the antioxidant astaxanthin merits further investigation as a potential stress protectant for improvement of tolerance in yeast strains.
We thank Ajinomoto Co. (Tokyo, Japan) for providing X. dendrorhous strains and C. Schmidt-Dannert (University of Minnesota, St. Paul) for providing pUCmod-crtZ and helpful discussion. We also thank A. Takehara, T. Katsuragi, T. Kaino, and I. Ohtsu of our laboratory for their discussions and technical assistance with this work.
This work was supported by a grant from the Nissui Research Foundation to H.T.
Published ahead of print on 2 October 2009.