Ochratoxin A Production and Amplified Fragment Length Polymorphism Analysis of Aspergillus carbonarius, Aspergillus tubingensis, and Aspergillus niger Strains Isolated from Grapes in Italy

ABSTRACT Ochratoxin A is a potent nephrotoxin and a possible human carcinogen that can contaminate various agricultural products, including grapes and wine. The capabilities of species other than Aspergillus carbonarius within Aspergillus section Nigri to produce ochratoxin A from grapes are uncertain, since strain identification is based primarily on morphological traits. We used amplified fragment length polymorphisms (AFLPs) and genomic DNA sequences (rRNA, calmodulin, and β-tubulin genes) to identify 77 black aspergilli isolated from grape berries collected in a 2-year survey in 16 vineyards throughout Italy. Four main clusters were distinguished, and they shared an AFLP similarity of <25%. Twenty-two of 23 strains of A. carbonarius produced ochratoxin A (6 to 7,500 μg/liter), 5 of 20 strains of A. tubingensis produced ochratoxin A (4 to 130 μg/liter), 3 of 15 strains of A. niger produced ochratoxin A (250 to 360 μg/liter), and none of the 19 strains of Aspergillus “uniseriate” produced ochratoxin A above the level of detection (4 μg/liter). These findings indicate that A. tubingensis is able to produce ochratoxin and that, together with A. carbonarius and A. niger, it may be responsible for the ochratoxin contamination of wine in Italy.

Ochratoxin A (OTA) is an important mycotoxin; is considered to be nephrotoxic, immunotoxic, genotoxic, and teratogenic; and has been classified by the International Agency for Research on Cancer as a possible human carcinogen (group 2B) (10). Ochratoxin A is produced by a small number of species in the genera Aspergillus, Petromyces, Neopetromyces, and Penicillium (14) and can contaminate various agricultural products, including grapes and wine (13,34,36). Accurate identification of ochratoxigenic fungi is of great importance because the toxin profiles of individual species vary and because the fungi that are present limit and define the potential toxicological risks (40). Unfortunately, the taxonomy of Aspergillus section Nigri is not completely resolved, especially within the Aspergillus niger aggregate (5,20,35,38). The A. niger aggregate of Al Musallam (5) is currently described as two species, A. foetidus and A. niger, that are subdivided further into seven varieties, based on morphological and cultural criteria (5,18,38). Molecular studies support the division of the A. niger aggregate into two morphologically indistinguishable species, A. niger and A. tubingensis (20,32,48).
The presence of ochratoxin A in wine is a relatively recent mycotoxicological problem (31,33,44,49) that is due to contamination by black aspergilli, primarily strains of A. carbonarius and others belonging to the A. niger species aggregate (8,21,46). These reports are all based on morphological identi-fications, which have limited ability to distinguish species in the A. niger aggregate. Amplified fragment length polymorphism (AFLP) analysis, described by Vos et al. (50), can be used for strain identification, especially at low taxonomic ranks (41,42). Recent studies suggest that these markers can be used to evaluate genetic relatedness among fungal species (23,51) and to clarify relationships within or between closely related groups or species (24,47). The advantages of this technique are its high discriminatory power, reproducibility, and robustness. It also can be easily automated (4,45), and numerous independent polymorphisms can be identified with relatively little change in the protocol.
Our objective in this study was to determine which strain types of the black aspergilli isolated from grapes in Italy, characterized by different molecular approaches (AFLP analysis and DNA sequencing of 28S, internal transcribed spacer [ITS], calmodulin, and ␤-tubulin genes), can produce ochratoxin A. This report is the first to combine a toxicological characterization of black aspergilli isolated from grapes with different molecular techniques of genetic identification. Field sampling and fungal isolation. Aspergilli were selected during a survey carried out in 2000 and 2001 in 16 vineyards. Five grape berries were taken from a single bunch of grapes collected from each of 10 plants along a diagonal transect in each vineyard. The berries were placed in moist chambers, i.e., petri dishes (9 cm in diameter) containing disks of blotting paper (8 cm in diameter) wetted with 2 ml of sterile water, that were then sealed with Parafilm and incubated at 25 Ϯ 2°C for 7 days. Fungal colonies were transferred to standard identification media, Czapek agar, Czapek yeast agar, and malt extract agar (39); incubated at 25 Ϯ 2°C in the dark for 7 days; and identified according to standard morphological criteria (17). Seventy-seven strains derived from single-conidialhead subcultures were assigned ITEM numbers and deposited in the culture collection of the Institute of Sciences of Food Production (http://www.ispa.cnr .it/Collection). AFLP analysis. Fungal strains were grown in shake cultures (150 rpm; 25°C; 2 days) in Wickerham's medium (glucose, 40 g; peptone, 5 g; yeast extract, 3 g; malt extract, 3 g; and distilled water to 1 liter). Genomic DNA was extracted with the E.Z.N.A. Fungal DNA Miniprep Kit (Omega Bio-tek, Doraville, GA) according to the manufacturer's protocol. DNA was dissolved in sterile water, diluted to 20 ng/l, and stored at Ϫ20°C.

Strains
We used the AFLP Microbial Fingerprinting kit (Applied Biosystems-Perkin-Elmer Corporation, Foster City, CA) according to the manufacturer's instructions. Approximately 10 ng of genomic DNA from each isolate was cut with EcoRI and MseI (New England Biolabs, Hitchin, Hertfordshire, United Kingdom), and the DNA fragments were ligated to double-stranded restriction sitespecific adaptors from the kit. A preselective PCR (72°C for 2 min; 20 cycles of 94°C for 20 s, 56°C for 30 s, and 72°C for 2 min; and then holding at 4°C) was carried out in a 20-l (final volume) mixture. For the selective PCR, 1.5 l of a 1:20 dilution of the first PCR product was amplified in a 10-l (final volume) mixture using selective primers. Four separate primer combinations were utilized for the selective amplification: EcoRIϩAC and MseIϩCC; EcoRIϩAT and MseIϩCG; EcoRIϩAC and MseIϩCA; and EcoRIϩG and MseIϩCT. The EcoRI primers were labeled with fluorescent dye (Applied Biosystems). The PCR program for selective AFLP amplification was one cycle of 94°C for 2 min and one cycle of 94°C for 20 s, 66°C for 30 s, and 72°C for 2 min; this cycle was followed by nine cycles in which the annealing temperature was lowered each cycle by 1°C from 65°C to 57°C. After that, 20 cycles of 94°C for 20 s, 56°C for 30 s, and 72°C for 2 min were performed, followed by a final extension step at 60°C for 30 min, and then holding at 4°C indefinitely by using a model 9700 GeneAmp PCR system.
After amplification, 1 l of reaction product was mixed with 20 l formamide and 0.5 l GeneScan-500 (ROX) size standard (Applied Biosystems), ranging from 35 to 500 bp in length. The mixture was heated for 2 min at 95°C and snap cooled on ice. The product was separated by capillary electrophoresis on an ABI PRISM 310 Genetic Analyzer (Applied Biosystems). After electrophoresis, the pattern was extracted with GeneScan collection software version 3.1.2 (Applied Biosystems), and the fingerprints were analyzed with Genotyper software (Applied Biosystems). DNA samples from five strains were tested in triplicate, and DNA samples from the other strains were tested in duplicate. DNAs from three replicate cultures of five strains each were also tested.
Peak height thresholds were set at 200. The Genotyper software (Applied Biosystems) was set to medium smoothing. Bands of the same size in different individuals were assumed to be homologous and to represent the same allele. Bands of different sizes were treated as independent loci with two alleles (present and absent). Data were analyzed with an AFLP manager database developed by ACGT BioInformatica S.r.l. (Bari, Italy) and were exported in a binary format with "1" for the presence of a band/peak and "0" for its absence. For clustering, two different analyses were performed. Fragments between 100 and 500 bp and between 200 and 500 bp were analyzed with NTSYS software by using the Dice similarity coefficient and clustered by the unweighted pair group method (27); the clusters obtained were also found by neighbor-joining analysis and maximumparsimony networks using MEGA version 3.0, with a bootstrap analysis of 1,000 repetitions (19).
Fungal-DNA amplification and sequencing. The identities of the ochratoxigenic species (A. carbonarius, A. niger, and A. tubingensis) also were confirmed by DNA sequencing. The nuclear rRNA gene containing the ITS region and domains D1 and D2 at the 5Ј end of the 28S rRNA gene were amplified with primers F65/R635 and ITS5/ITS4, respectively (28). The calmodulin and ␤-tubulin genes were amplified by using the conditions and primers described by O'Donnell et al. (29).
After amplification, the PCR products were purified by agarose gel electrophoresis and excised from the agarose gel using spin columns (DNA Gel Extraction Kit; Millipore Corporation, Bedford, MA). Sequence analysis was performed with the Big Dye Terminator Cycle Sequencing Ready Reaction Kit for both strands, and the sequences were aligned with the MT Navigator software (Applied Biosystems). All the sequencing reaction mixtures were purified by gel filtration through Sephadex G-50 (Amersham Pharmacia Biotech, Piscataway, NJ) equilibrated in double-distilled water and analyzed on the ABI PRISM 310 Genetic Analyzer (Applied Biosystems). The resulting sequences of all the isolates were aligned by the Clustal method with the DNAMAN program (Lynnon Corporation, Quebec, Canada).
Ochratoxin A production. Strains were grown in triplicate in 100-ml stationary cultures of enriched Czapek yeast broth (10) (6) and stored at Ϫ20°C. Working standards were prepared by evaporating an exact volume of the calibrated solution under a stream of nitrogen and redissolving the residue in the mobile phase.
The HPLC system consisted of a Perkin-Elmer 200 instrument equipped with an ISS 200 sampling system (loop volume, 150 l) and a Jasco FP-920 fluorescence detector set at 333-nm excitation and 470-nm emission. The system was controlled by Perkin-Elmer Turbochrom PC software. A Select B RP-8 column (5-m particle size; 150-by 4-mm inside diameter; Merck, Darmstadt, Germany) was employed at ambient temperature (20 to 25°C), with a mobile phase of acetonitrile-2% acetic acid (41:59 for ochratoxin A and 55:45 for ochratoxin A methyl ester) at 1.2 ml/min. The injection volume was 30 l. Ochratoxin A standards of 2 to 60 pg were injected. Peak areas were quantified with the Turbochrom PC software. Ochratoxin A in the extracts was methylated, and the extracts were reanalyzed by HPLC for qualitative confirmation of positive samples (52). The detection limit was 4 g/liter. All analyses were run in triplicate, and the mean values were reported; standard deviations were Ͻ5% of the reported values.
Nucleotide sequence accession numbers. The calmodulin sequences were deposited in the EMBL nucleotide sequence database (Table 1).

RESULTS
We examined 77 Aspergillus strains ( Table 2) selected arbitrarily from 692 isolates of black aspergilli isolated from grapes in Italy during a 2-year survey. Clear polymorphisms both within and between species were obtained by AFLP analysis for each of the four primer pairs. Each primer combination consistently distinguished the taxa in section Nigri, as evidenced by the distribution of the ex-type strains in the neighbor-joining dendrogram (Fig. 1). Unweighted pair group method and maximum-parsimony analyses (data not shown) also divided the isolates into the same clusters, with a bootstrap support of 89 to 100% and 100%, respectively. The 77 strains analyzed clearly separated into four clusters: A. tubingensis, Aspergillus "uniseriate," A. carbonarius, and A. niger (Fig. 1), with a similarity of Ͻ20% for strains in different clusters. Strains were assigned to a species if they shared more than 50% of the bands present in an ex-type strain. All strains of A. carbonarius and A. tubingensis had high similarity (more VOL. 72, 2006 OCHRATOXIN A AND BLACK ASPERGILLI FROM GRAPES 681 than 50%) with respect to their ex-type strains. The A. niger cluster strains were more similar to the ex-type strains of A. awamori (58%) than to the A. niger ex-type strain (40 to 45%). The Aspergillus "uniseriate" strains from grapes were at best distantly related to the ex-type strains of A. japonicus and A. aculeatus, with a similarity of Ͼ20%. We termed this cluster Aspergillus "uniseriate" because we could not assign the strains to any of the presently accepted uniseriate taxa in the Nigri section (Fig. 1). Both the AFLP analyses at fragment cutoffs of 100 and 200 bp gave the same grouping and similarity among the four clusters and their related ex-type strains. Aspergillus carbonarius. The A. carbonarius cluster contained all 23 strains identified morphologically as A. carbonarius. All the isolates clustered at a similarity of 68 to 78% with the A. carbonarius ex-type, CBS 111.26 (Fig. 1), while, among them, most of the members of the population shared Ն78% of the bands. Aspergillus carbonarius was the most toxigenic species, with 22/23 strains positive for ochratoxin A production and with ITEM 5001 the only strain in the species that did not produce OTA. The amount of ochratoxin A produced varied widely by strain, from 6 g/liter by ITEM 4864 to 7,500 g/liter by ITEM 5010 (  Aspergillus niger. Fifteen strains were grouped in the A. niger cluster with an AFLP similarity of 40 to 46% to the ex-type of A. niger, CBS 554.65, and with a similarity of 58 to 75% to the ex-type strain of A. awamori, CBS 557.65, also identified as A. niger by ␤-tubulin analysis (39). However, the AFLP analysis at a cutoff of 200 bp revealed a greater distance between the ex-type of A. niger, CBS 554.65, and the 15 strains (35% similarity). The sequences of the rRNA genes for the ex-type strains and the 15 field strains in this group were identical. The calmodulin and ␤-tubulin genes of the 15 strains had identities of 98% and 100% with those from the A. niger and A. awamori type strains, respectively. Three of the 15 strains in this group produced detectable ochratoxin A (250 to 360 g/liter) (

DISCUSSION
Aspergillus carbonarius contained both the highest proportion of toxigenic strains (95%) and the highest producers of ochratoxin A (up to 7.5 g/ml). Previous reports of toxin production by A. carbonarius found that 25 to 100% of the strains examined produced ochratoxin A in amounts ranging from 0.3 to 234 g/g when cultured in an agar medium (12,37). Further study is needed to determine the cause of this variation. A. carbonarius is a source of ochratoxin A in crops such as coffee and dried vine fruits (1,16), but its incidence probably is underreported, since all of the black aspergilli are commonly reported only as "A. niger" (2). This is the first report of ochratoxin A production by molecularly characterized strains of A. carbonarius from grapes. Although A. carbonarius has distinctive morphological traits that are easily identified in a microscopic evaluation, molecular tools can speed the identification process and make it more objective. In this respect, AFLP analysis also can be used for identification purposes (42) and for the development of species-specific PCR primers based on unique AFLP fragments (43).
The other ochratoxin-A-producing Aspergillus species we found were A. tubingensis and A. niger. These species are difficult to differentiate on the basis of morphology, with reliable characteristics for distinguishing A. niger, A. foetidus, and A. tubingensis remaining to be identified (40). Aspergillus tubingensis was first described by Mosseray (25) (2,20,32,40), however, strongly support our results, which are consistent with the hypothesis that A. tubingensis is clearly distinct from A. niger. Of special importance is our determination that A. tubingensis can produce ochratoxin A. It is not clear what proportion of the strains of "A. niger" previously reported to produce this mycotoxin and recovered from coffee beans (26), vine fruits (15,21), and animal feed (11), or available from culture collections (3,30,46), are A. tubingensis. The identity of the strains from grapes as A. tubingensis is consistent with DNA sequence comparisons made with four previously analyzed diagnostic genes (3,20,32,40). In contrast with our results, Accensi et al.
(3) evaluated isolates of the A. niger aggregate (none from grapes) and found that "N-type" A. niger strains produced ochratoxin A but that A. tubingensis strains (termed the "T type") could not. The strains of A. tubingensis and A. niger differ from those of A. carbonarius both in the percentages of ochratoxin A-producing isolates recovered (25%, 20%, and 95%, respectively) and in the amounts of ochratoxin produced in liquid cultures (4 to 130, 255 to 357, and 6 to 7,500 g/liter, respectively). These results confirm that A. carbonarius is an important ochratoxinproducing fungus in wine and that, together with A. tubingensis and A. niger, it may be responsible for the ochratoxin A contamination of wine in Italy.
Although two recent studies (9,11) reported the production of ochratoxin A by A. japonicus strains isolated from grapes and animal feed, none of our Aspergillus "uniseriate" strains produced detectable ochratoxin A. The genetic distance, based on AFLPs, of our strains from the ex-type strains, however, leaves the taxonomic status of our cultures in need of further definition and resolution, as they could represent a new population or an as-yet-undescribed Aspergillus species.
AFLP analysis is a useful and powerful tool for studying the genetic diversity of black aspergilli. Our AFLP results, combined with DNA sequence analysis of diagnostic genes, confirm the validity and utility of the AFLP technique for evaluating genetic relatedness among fungal species (23,51), and in particular, for resolving relationships within or between closely related groups or species (24,47). Previous studies have shown that this technique is suitable for differentiating fungal strains at and below the species level (7,22,41,42). Our study suggests that AFLPs have utility for this purpose in this group of Aspergillus species as well. A combination of biochemical and molecular methods is needed to correctly evaluate the potential toxicological risk in grapes caused by these fungi.