A Rapid Assay to Detect Toxigenic Penicillium spp. Contamination in Wine and Musts

Wine and fermenting musts are grape products widely consumed worldwide. Since the presence of mycotoxin-producing fungi may greatly compromise their quality characteristics and safety, there is an increasing need for relatively rapid “user friendly” quantitative assays to detect fungal contamination both in grapes delivered to wineries and in final products. Although other fungi are most frequently involved in grape deterioration, secondary infections by Penicillium spp. are quite common, especially in cool areas with high humidity and in wines obtained by partially dried grapes. In this work, a single-tube nested real-time PCR approach—successfully applied to hazelnut and peanut allergen detection—was tested for the first time to trace Penicillium spp. in musts and wines. The method consisted of two sets of primers specifically designed to target the β-tubulin gene, to be simultaneously applied with the aim of lowering the detection limit of conventional real-time PCR. The assay was able to detect up to 1 fg of Penicillium DNA. As confirmation, patulin content of representative samples was determined. Most of analyzed wines/musts returned contaminated results at >50 ppb and a 76% accordance with molecular assay was observed. Although further large-scale trials are needed, these results encourage the use of the newly developed method in the pre-screening of fresh and processed grapes for the presence of Penicillium DNA before the evaluation of related toxins.


Introduction
Wine is one of the major processed grape (Vitis vinifera L.) products, with a worldwide production of 26,404,435 tons [1], obtained by the total or partial alcoholic fermentation of grapes or musts [2]. Usually, red wines are produced from black grape musts, and fermentation occurs in presence of the grape skins, whereas white wines are produced by fermentation of the juice obtained by pressing crushed grapes. The process stops either naturally, when sugars are completely converted, or artificially, by lowering the temperature. Musts can also undergo "enrichment"-that is, an increase in the sugar concentration prior to fermentation-to gain a proper final level of alcohol in the wine. However, fermenting musts are not only an intermediate product, as they are directly consumed in wine-growing areas of Northern Europe (mainly Germany and Austria) during the autumn season [3], in particular by children [4]. Their overall quality is usually poor, as they represent the wastes of the production of quality-tested wine. Therefore, the risk of contamination by toxic metabolites produced by

Set up of Experimental Design
Two sets of primer pairs designed upon a portion of β-tubulin gene, with different annealing temperatures, were used to detect the presence of Penicillium spp. in extracted samples. The first set of primers (NESF-NESR) generating PCR fragments of 320 bp worked as the "outer" primers to delineate the chosen target sequence (Figure 1). This primer pair was selected to hybridize Toxins 2016, 8,235 3 of 12 at a higher temperature (60˝C), conferring selectivity to the reaction. The second set of primers, HRMF-HRMR, producing PCR fragments of 96 bp, was defined to act as "inner" primers at lower hybridization temperatures (55˝C). In order to perform the single-tube nested real-time PCR approach, two independent temperature phases were established. During phase 1, PCR fragments of 320 bp were amplified, to serve as DNA template later on in the reaction, with no fluorescence acquisition. Phase 2 was planned to obtain PCR fragments of 96 bp using the 320 bp fragments as template, and the collection of fluorescence was performed at the end of each cycle. The number of cycles to be used in each phase was selected according to the best performance in nested qPCR trials: phase 1 was set at 15 cycles, whereas phase 2-with fluorescence signal acquisition-was carried out using 30 cycles. Two sets of primer pairs designed upon a portion of β-tubulin gene, with different annealing temperatures, were used to detect the presence of Penicillium spp. in extracted samples. The first set of primers (NESF-NESR) generating PCR fragments of 320 bp worked as the "outer" primers to delineate the chosen target sequence (Figure 1). This primer pair was selected to hybridize at a higher temperature (60 °C), conferring selectivity to the reaction. The second set of primers, HRMF-HRMR, producing PCR fragments of 96 bp, was defined to act as "inner" primers at lower hybridization temperatures (55 °C). In order to perform the single-tube nested real-time PCR approach, two independent temperature phases were established. During phase 1, PCR fragments of 320 bp were amplified, to serve as DNA template later on in the reaction, with no fluorescence acquisition. Phase 2 was planned to obtain PCR fragments of 96 bp using the 320 bp fragments as template, and the collection of fluorescence was performed at the end of each cycle. The number of cycles to be used in each phase was selected according to the best performance in nested qPCR trials: phase 1 was set at 15 cycles, whereas phase 2-with fluorescence signal acquisition-was carried out using 30 cycles.

Specificity and Sensitivity Assay
In BLAST analyses, Penicillium primer sets did not match any of the available DNA sequences in international databases other than their reference genus. Moreover, specificity tests were conducted amplifying DNA from different fungal genera and species commonly associated to grape (Table 1). A positive amplification (increase of fluorescence) was obtained by the sole Penicillium strains. No cross-amplification with grape (Vitis vinifera) DNA was observed. Table 1. Results of nested real-time PCR amplifications of gene applied to fungal genera more frequently reported on grapes. Penicillium spp. and grape DNA were included as controls.

Specificity and Sensitivity Assay
In BLAST analyses, Penicillium primer sets did not match any of the available DNA sequences in international databases other than their reference genus. Moreover, specificity tests were conducted amplifying DNA from different fungal genera and species commonly associated to grape (Table 1). A positive amplification (increase of fluorescence) was obtained by the sole Penicillium strains. No cross-amplification with grape (Vitis vinifera) DNA was observed. Table 1. Results of nested real-time PCR amplifications of gene applied to fungal genera more frequently reported on grapes. Penicillium spp. and grape DNA were included as controls.

Pex6
Penicillium expansum + Pex29 Penicillium chrysogenum + Pex30 Penicillium crustosum Vitis vinifera -To evaluate the sensitivity of the reaction and to quantify Penicillium DNA, a standard curve was drawn ( Figure 2). Five 10-fold dilutions in the range 100-0.001 pg/µL of P. expansum DNA were amplified. The standard curve showed a linear correlation (p ď 0.001) between input DNA and Ct values, with R 2 = 0.9961. The system was able to efficiently amplify up to 1 fg of target DNA. amplified. The standard curve showed a linear correlation (p ≤ 0.001) between input DNA and Ct values, with R 2 = 0.9961. The system was able to efficiently amplify up to 1 fg of target DNA.
In order to evaluate the influence of grape extracts on the quantification of fungal DNA, the experiment was repeated, adding grape DNA to all serial dilutions. The obtained curve was not influenced by the presence of grape DNA, since an identical detection limit and very similar determination coefficient (R 2 = 0.9653) were observed ( Figure 2).

Penicillium Detection in Real Samples
Eighty-two musts and wines (whites and reds) were collected from private wineries in Southern Italy. They came from tanks (large resin-coated cement underground containers), cisterns (large circular stainless steel vessels on legs), and silos (small cisterns). Samples underwent DNA extraction, and, in order to prevent false-negatives, their suitability to PCR amplification was confirmed using grape-specific primers. Of the analysed samples, 38 (46%)-made up of 19 musts (6 whites and 13 reds) and 19 wines (7 whites and 12 reds)-were found positive for Penicillium contamination (Table  2). In particular, they came from 18 (out of 31, 58%) tanks, 5 (out of 22, 23%) cisterns, and 15 (out of 28, 54%) silos. Therefore, there was a significantly lower frequency of contamination among samples coming from cisterns (Kruskal-Wallis, p < 0.05). Penicillium DNA was found in the range 0.001-2.634 pg/μL, with the red must SS26 and white wine T40 containing the higher and lower quantity of Penicillium DNA, respectively. However, there were no significant differences between musts/wines and reds/whites.  In order to evaluate the influence of grape extracts on the quantification of fungal DNA, the experiment was repeated, adding grape DNA to all serial dilutions. The obtained curve was not influenced by the presence of grape DNA, since an identical detection limit and very similar determination coefficient (R 2 = 0.9653) were observed ( Figure 2).

Penicillium Detection in Real Samples
Eighty-two musts and wines (whites and reds) were collected from private wineries in Southern Italy. They came from tanks (large resin-coated cement underground containers), cisterns (large circular stainless steel vessels on legs), and silos (small cisterns). Samples underwent DNA extraction, and, in order to prevent false-negatives, their suitability to PCR amplification was confirmed using grape-specific primers. Of the analysed samples, 38 (46%)-made up of 19 musts (6 whites and 13 reds) and 19 wines (7 whites and 12 reds)-were found positive for Penicillium contamination ( Table 2). In particular, they came from 18 (out of 31, 58%) tanks, 5 (out of 22, 23%) cisterns, and 15 (out of 28, 54%) silos. Therefore, there was a significantly lower frequency of contamination among samples coming from cisterns (Kruskal-Wallis, p < 0.05). Penicillium DNA was found in the range 0.001-2.634 pg/µL, with the red must SS26 and white wine T40 containing the higher and lower quantity of Penicillium DNA, respectively. However, there were no significant differences between musts/wines and reds/whites. White must Cistern -C12a Red must Cistern 0.010 C12b White must Cistern -

Patulin Quantification in Real Samples
Patulin occurrence and concentration was estimated for the seventeen samples of red and white musts and wines that resulted positive for the presence of Penicillium (Table 3). Thirteen of the analysed samples resulted contaminated in the range 27-1911 µg/L, with white wines T17 and SS21 as the most and least contaminated samples, respectively. There were no significant differences in terms of toxin contamination between musts and wines, or reds and whites. A concordance between presence/absence of the fungus and of the toxin was observed for 13 samples (76%), whereas in four samples, Penicillium but not patulin was detected. There was no linear correlation between Penicillium DNA and patulin contamination extents.

Discussion
Penicillium species are ubiquitous fungi associated with organic matter in nature. Although mainly linked to other commodities, their presence as epiphytes on grapes has been reported, with the frequency increasing considerably as berries mature [6]. However, species of Penicillium are gaining attention not only as grapevine pathogens at harvest [23], but also during the postharvest phase and winemaking [9,24].
Fungi cause drastic chemical and enzymatic modifications depending on grape variety and production stage [25,26], leading to serious sensory defects and risks of contamination by toxic metabolites (including patulin) in wine. Consequently, there is an increasing interest in determining contamination by Penicillium spp. of grapes, musts, and wines-especially those obtained from partially dried grapes. For example, the withering process for the production of passito wines (e.g., Amarone, Sfurzat, Vin Santo, Recioto) lasts up to 5 months in specific thermo-hygrometric conditions, in which fungal contamination can take place [27].
The correct evaluation of the potential presence of pathogens/metabolites using molecular assays is highly dependent on numerous factors, such as the type of food matrix, the mycotoxin/DNA markers, and the chosen methodology, among others [28]. In this work, we present an alternative method based on the assembly of two DNA-based techniques (nested PCR and real-time PCR) for the detection of Penicillium DNA in wines and musts. The task was to set up an assay that is easily applicable on a large number of samples at once, thus representing a quick and efficient pre-screening before traditional chemical analyses.
Regarding the nested real-time PCR assay developed in this work, our system was able to detect the presence of Penicillium in 46% of the tested samples, with samples coming from cisterns showing the lowest contamination. The detection limit was 1 fg, a result particularly interesting, considering that the average weight of the haploid genome of Penicillium spp. is reported to be 31 fg [29]. Moreover, this sensitivity level is much better than levels reported in literature concerning Penicillium detection in food matrices [17][18][19]. Indeed, by the introduction of the nested approach, it was possible to enhance the performance of a traditional real-time PCR assay. The proposed new detection system presents the advantage of high specificity conferred by the use of two pairs of primers at different annealing temperatures. In particular, the empirical rule for single-tube nested real-time PCR system-Ta (inner primers) < Ta (outer primers)-used for the detection of Ara h 3 [20], hsp1 [21], and Pru du 6 [22] allergens, was followed. The single-tube nested real-time PCR approach presented high performance criteria and apparent robustness, since it was not affected by shifts in temperature, time, cycle number (despite the existence of two different reaction protocols within the same assay), or the coexistence of grape DNA. Moreover, the single-tube amplification could be particularly efficient in preventing the cross-contamination and false negative results that are the major drawbacks of a nested approach.
As confirmation of Penicillium contamination, patulin presence was evaluated in representative samples. The mycotoxin was found in 71% of analysed wines and musts. With one exception (SS21), it was above the EU regulatory limit of 50 µg/kg foreseen for fermented apple juices, since there are no specific regulatory limits for patulin in wines and musts. The huge amount of toxin recorded even in wines strongly evidences the risks for consumers' health, stating the need to detect and control the presence of patulin-producing fungi such as Penicillium all through the winemaking chain. The issue of the presence of patulin in grape musts was already addressed in Austria, as it was detected (maximum values 23.6-750 µg/kg) in 86 of the 164 samples surveyed from 1996 to 2000 [30]. This finding was alarming, considering that fresh grape must is offered to children as a non-fermented and unheated drink in the Austrian wine-growing regions [30].
A 76% accordance between molecular and toxicological data was recorded, although it was not quantitative. Similarly, Majerus et al. [3] found that contamination of grape must with patulin did not necessarily correlate with the moulding of the product, and Fredlund et al. [31] reported that the levels of both deoxynivalenol and zearalenone did not correlate with the DNA levels of Fusarium culmorum or other Fusarium species. In four samples containing Penicillium DNA, no patulin was detected. This was not surprising, since not all Penicillium species reported on grape are able to produce patulin [8]. Moreover, 60-plus species of moulds encompassing over 30 genera (including Paecilomyces, Saccharomyces, Alternaria, Byssochlamys, and Aspergillus) [32]-many of which have been reported on grape-produce patulin. In a recent study, the presence of patulin biosynthetic gene patN proved to be not predictive for patulin contamination [33]. Finally, it has to be considered that Penicillium produces several other toxic compounds (e.g., citrinin, chaetoglobosins, etc.) that can affect the quality of and safety of the product [34], and thus have to be monitored.

Conclusions
In conclusion, the single-tube nested real-time PCR method presented in this work constitutes an alternative, quick, and reliable approach for the detection of Penicillium even at trace levels in grape-derived products. The interesting results obtained with this approach highlight the usefulness of this new tool and its potential for the identification of pathogens in food matrices, for which further research work is needed. Moreover, the high patulin levels found in analyzed samples suggest the need to pay for greater attention to Penicillium toxins in musts and wines.

Sample Collection
During autumn 2013 and spring 2014, 82 musts and wines (whites and reds) were collected from private local wineries in the Apulia region, Southern Italy (Table 1). After 10 min of stirring, 6 L of each sample were collected, divided in three bottles of 2 L each, and stored at 4˝C until use. Among them, 17 samples were analyzed for patulin content.

DNA Extraction
DNA extraction from musts and wines was performed according to the method of di Rienzo et al. [35]. The DNA was further purified using the HiYield™ Gel/PCR Fragments Extraction Kit (Real Genomics, Banqiao City, Taiwan) according to manufacturer instructions, performing two washing steps and recovering the DNA with the elution buffer pre-heated at 60˝C. The DNA concentration, purity, and integrity were determined both by the Nano-Drop™ 2000 Spectrophotometer (Thermo Scientific, Waltham, MA, USA) and electrophoresis on a 0.8% agarose Tris/Borate/EDTA (TBE) gel. In order to prevent false negatives, the suitability of extracted DNA to PCR amplification was evaluated using V. vinifera primers [36].

Nested One-Tube Real-Time PCR Assays
Real-time PCR assays were performed in 10 µL of total reaction volume. For each reaction tube, 5 µL of DNA, 1ˆSYBR ® Select Master Mix (Thermo Scientific), 300 nM of each inner primer HRMF1/HRMR1, and further 300 nM of each outer primer NESF/NESR were used. All real-time PCR assays were made on an iCycler iQ thermal cycler (BioRad, Hercules, CA, USA).
Nested real-time PCR assays were carried out with two different temperature programs: phase 1, performed without collecting fluorescence signal; and phase 2, with collection of the fluorescence signal at the end of each cycle. The number of cycles used in each phase was defined as follows: phase 1 from 5 to 15 cycles; phase 2 from 30 to 40 cycles. The following temperature protocol was used: 50˝C for 2 min, 95˝C for 2 min, 5-15 cycles at 95˝C for 15 s, 60˝C for 15 s (phase 1), and 30-40 cycles at 95˝C for 15 s, 55˝C for 15 s, and 72˝C for 15 s (phase 2). Fluorescence was acquired during the extension at 72˝C to further improve specificity and signal-to-noise ratio [37]. Data were collected and analyzed using the iCycler iQTM associated software (Real time Detection System Software, version 3.0, BioRad). Cycle threshold (Ct) values were calculated using the software at automatic threshold setting.

Specificity and Sensitivity Assay
To test the specificity of the reaction, the DNA of the most frequent fungal genera reported on grape, plus the DNA of grape and of Penicillium spp., was amplified as reported above.
Moreover, to assess the sensitivity of the assay, Penicillium DNA was serially diluted ten-fold with sterile water to yield final concentrations from 100 to 0.001 pg/µL, and amplified as described above. A standard curve was generated by plotting the DNA amounts [log (pg)] against the corresponding Ct value. Determination coefficient (R 2 ) and linear equation were calculated. In order to evaluate the influence of co-extracted DNA on the efficiency of the two primer sets, a standard curve was drawn by adding 50 ng of V. vinifera DNA to each reaction mixture. Penicillium concentration in unknown samples was extrapolated from the standard curve.

Patulin Evaluation
For confirmation, the presence and concentration of patulin was evaluated in 17 samples positive to molecular assays.

Chemicals and Reagents
All reagents had a purity >98.0% and were purchased from (Sigma Aldrich, Milan, Italy). A patulin stock solution in methanol was prepared at a concentration of 607.6 mg/L and stored at´20˝C. A working solution of 6.08 mg/L was also prepared.

Chromatographic and Mass Spectrometric Conditions
Analyses were performed by a Liquid Chromatograph NEXERA X2 LC30AD System (Shimadzu, Milan, Italy), equipped with a binary solvent delivery system, degasser, autosampler, and column heater. The separation was performed on a LUNA C8 analytical column (150 mmˆ2 mm I.D.), with 5 µm particles, from Phenomenex (Torrance, CA, USA). The detection system was an AB SCIEX LC/MS/MS Triple Quad 5500 System tandem mass spectrometer equipped with an electrospray ionization interface (ESI) operating in the negative ion mode, using multiple reaction monitoring (MRM). A gradient elution was performed using a mobile phase (flow rate 0.25 mL/min) constituted by water (1% CH 3 COOH and 5 mM C 2 H 3 O 2 NH 4 ) and methanol (1% CH 3 COOH and 5 mM C 2 H 3 O 2 NH 4 ), eluent A and B, respectively. The program started at 10% eluent B and ramped to 40% at 5 min and to 90% at 11 min. It remained constant for 4 min and then decreased linearly to 10% of eluent B. This condition was kept constant for 5 min, and the column was re-equilibrated to the initial mobile phase composition. The column temperature was kept at 40˝C. The mass spectrometer ion source parameters applied were: Curtain Gas 30.00 psi; Desolvation Gas Temperature 550.00˝C; GS1 (air) 60.00 psi; GS2 (air) 55.00 psi; Ion Spray´4500.00 V. Collision energy and cone voltage acquisition parameters are reported in Table 4. The instrument had a limit of detection (LOD) of 0.02 mg/L and a limit of quantitation (LOQ) of 0.05 mg/L. The recovery of the method was in the range 81%-92%. Unknown samples were analyzed comparing standard patulin retention time and ion ratio (within 20%); quantification was performed by a six-point calibration curve (y = 5479.74x + 554.75, R 2 > 0.99) obtained for the mass fragment 152.9 Ñ 108.9. Table 4. Optimization of the collision energy and cone voltage for patulin by infusion of the mycotoxin directly into the LC effluent, and final acquisition parameters.

Statistical Analysis
Data processing and correlation analyses were performed using the statistical software package Statistics for Windows (StatSoft, Tulsa, OK, USA). Values were tested independently for normality using the Shapiro-Wilk (SW) test. Given that samples did not come from normally-distributed populations (SW test, p < 0.01), nonparametric tests were chosen for downstream analyses. The two-ways Wilcoxon (W) test and the Kruskal-Wallis (KW) test were applied to compare samples from two and three or more classes, respectively.