Cyclopiazonic acid in soft-ripened and blue cheeses marketed in the USA

ABSTRACT Strains of Penicillium camemberti and P. roqueforti are used in the production of soft-ripened and blue-veined cheeses. However, some strains can produce toxic secondary metabolites (mycotoxins), including α-cyclopiazonic acid (CPA), a neurotoxin. Data on the levels of CPA in cheeses marketed in the USA are extremely limited. An enzyme-linked immunosorbent assay was adapted for measuring CPA in soft-ripened and blue-veined cheeses. Recoveries from cheese curds were 103 ± 27% (n = 30). A total of 254 samples of soft-ripened, blue and miscellaneous cheeses were examined. CPA was detected in 36/79 (45.6%) of soft-ripened cheeses and in 41/168 (24.4%) of blue-veined cheeses. Median levels in positive samples were 48.5 µg/kg and 30 µg/kg, respectively. The highest levels found were 3,820 µg/kg (in a Brie), 1,250 µg/kg (in a blue) and 7,900 µg/kg (in a Monte Enebro). The implication of such exposures is unknown, as a consensus on acceptable intake remains to be established.


Introduction
Fungi are well known to cause spoilage of commodities at multiple points between the field and the table. However, fungi are also deliberately added to certain foods in order to add desirable flavours and textures (McSweeney 2017). Examples include the mouldripened cheeses. Most mould-ripened cheeses belong to one of the two groups: "soft ripened" cheeses and "blue-veined" cheeses. The former include the Camembert and Brie styles, which are inoculated or smeared with Penicillium camemberti. The latter include Roquefort, Stilton, Gorgonzola, DanaBlu and are inoculated with P. roqueforti. The Cambozola-style cheeses are ripened with both organisms. While the presence of fungi in these cheeses is desirable, certain fungi are capable of producing mycotoxins: secondary metabolites that are toxic to animals. Previous investigations have demonstrated the presence of neurotoxins in cheeses ripened with P. camemberti and P. roqueforti. One such toxin is α-cyclopiazonic acid. CPA, which was discovered over 50 years ago (Holzapfel 1968), belongs to a very diverse group of mycotoxins derived from tryptophan (Chang and Ehrlich 2011). Figure 1 shows related compounds like the speradines, aspergillines and cyclopiamides (Uka et al. 2017). CPA is produced by many species of Aspergili and Pencillium, including P. camemberti, P. griseofulvum, A. oryzae and A. flavus.
Aspergillus flavus is best known as a producer of aflatoxins, but many strains also produce CPA (Gallagher et al. 1978;Vinokurova et al. 2007).
Due to its low acute toxicity, CPA is generally considered less hazardous than the highly carcinogenic aflatoxins B 1 , G 1 , and M 1 . The comparison has perhaps diverted attention from the toxicity of CPA, which has shown adverse effects in multiple animal species, including ducklings, chickens, turkeys, pigs, dogs, monkeys, lactating ewes, Guinea pigs, rats, and mice (Ostry et al. 2018). Symptoms vary between species, but often include weight loss, diarrhoea, convulsions and degeneration of muscles and vicera (Voss 1990;Burdock and Flamm 2000;Chang et al. 2009). In young turkeys symptoms include lethargy, ataxia, drooped head and wings, ruffled feathers, regurgitation, anorexia, and decreased amount and loss of form of faeces (Miller et al. 2011). The major targets include the gastrointestinal tract, liver, kidney, and skeletal muscle. Poultry are well known to be affected by CPA, and this toxin is believed to have contributed to the original outbreak of Turkey X disease (Cole 1986). The possibility of a link between CPA and "Kodo poisoning" in humans was recently reviewed (Deepika et al. 2021) and requires further exploration. The proposed mechanism of action of CPA involves inhibition of the sarcoplasmic reticulum Ca 2+ ATPase and associated disruption of calcium homoeostasis (Riley et al. 1995). The daily intake deemed acceptable for humans was estimated to be 10 µg/kg body weight/day (Burdock and Flamm 2000), although a suggestion has been made that this value should be placed closer to 0.1 µg/kg bw/day (De Waal 2002).
As a result of the widespread distribution of CPAproducing organisms, the toxin has been found in a wide variety of foods. These include milk, cheese, maize, peanuts, figs, rice, tomato paste, wheat, and certain cured meats such as salami (summarised in Ostry et al. 2018). Several studies have focused on the presence of CPA in milk (Dorner et al. 1994;Prasongsidh et al. 1998;Losito et al. 2002;Oliveira et al. 2006Oliveira et al. , 2008 and, because P. camemberti is intentionally used in their ripening, the soft-ripened cheeses (Le Bars 1979;Finoli et al. 1999;Zambonin et al. 2001;Monaci et al. 2007;Ansari and Häubl 2016;Hossain et al. 2019). Efforts to monitor CPA in foodstuffs have resulted in the development of a wide range of analytical methods for detection. The toxin has an absorption in the ultraviolet (UV) region and can be rendered fluorescent after exposure to strong UV light (Maragos 2009). Liquid chromatographic (LC) methods have generally been based on either of these features or on mass spectrometry (Urano et al. 1992;da Motta and Valente Soares 2000;Hayashi and Yoshizawa 2005;Moldes-Anaya et al. 2009;Diaz et al. 2010;Soares et al. 2010;Ansari and Häubl 2016;Peromingo et al. 2018).
Several novel methods have been proposed to detect or predict CPA levels in cheeses. A predictive method based upon Fourier-transform mid-infrared attenuated total reflection spectroscopy (ATR/FTIR) was reported (Monaci et al. 2007). The technique combined surface measurement of mid-IR spectra with multi-variate statistics to predict CPA in samples of ewe cheeses. A biosensor based upon the technology of imaging surface plasmon resonance was also applied to the detection of CPA in cheeses (Hossain et al. 2019). Most recently, a lateral flow immunoassay employing a novel detection label (aggregation-induced emission luminogens) was developed though for the measurement of CPA in peanuts, rather than cheeses (Hu et al. 2021).
Previously our group developed two monoclonal antibodies (mAbs) and associated immunoassays for detecting CPA (Maragos et al. 2017). One of these assays was applied to the detection of CPA in samples of maize contaminated with A. flavus. To our knowledge, there have been no surveys published on the levels of CPA in cheeses marketed in the United States, other than a single report by Hossain et al. (2019) in which seven Camembert cheeses were examined. Data on occurrences in milk and cheese have been limited to Western Europe and in particular Italy. The objectives of this project were to adapt the previously developed ELISA for use with soft-ripened cheeses and to apply the method to the measurement of CPA in soft-ripened and blue-veined cheeses marketed in the USA.

Reagents
Except where noted otherwise, deionised water (Nanopure II; Thermo Fisher Scientific, Waltham, MA, USA) was used in the preparation of all reagents. The CPA was produced by MP Biomedicals, LLC (Solon, OH, USA). CPA stock solution was prepared at a nominal concentration of 2 mg/mL by dissolving solid toxin in LC grade acetonitrile. The actual concentration of CPA was determined by obtaining the UV spectrum of 1:200 dilutions of the stock in methanol using an Evolution 201 UV/Visible spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Spectra were collected over the range of 190 nm to 450 nm in 0.1 nm increments (1 nm bandwidth, 0.4 s integration time, speed 150 nm/min). The spectra were compared to published spectra to qualitatively confirm the purity of the stocks. Quantification of the stock solution was based upon the molar absorptivity of 20,417 at 284 nm (Holzapfel 1968;Nesheim and Stack 2001). The CPA mAb 1418 and the immobilised antigen CPA-β-lactoglobulin conjugate (CPA-BLG) were produced at the USDA-National Center for Agricultural Utilisation Research (USDA-NCAUR, Peoria, IL, USA), as described previously (Maragos et al. 2017). Buffers used in the immunoassays included (1) 10 mM phosphate buffered saline (PBS: 0.01 M sodium phosphate and 0.15 M sodium chloride in water, adjusted to pH 7.2); (2) Citrate phosphate buffer (CPB: 0.05 M sodium citrate, 0.1 M sodium phosphate, adjusted to pH 5.0); (3) Tween-PBS (0.02% v/v Tween-20 in 0.01 M PBS pH 7); and (4) PVA-PBS (1% w/v Polyvinyl alcohol in 0.01 M PBS). PVA was purchased from Sigma-Aldrich (Milwaukee, WI, USA). Bovine serum albumin (BSA) was obtained from Thermo Fisher Scientific (Waltham, MA, USA). Czapek's broth medium (CZB) was purchased from Sigma-Aldrich (St. Louis, MO, USA). All other chemicals were reagent grade or better and purchased from major suppliers.

Sampling
Samples of soft-ripened cheeses (n = 79) and blueveined cheeses (n = 168) were purchased from groceries located in central Illinois (USA), central Wisconsin (USA), and online (iGourmet, West Pittston, Pennsylvania, USA). In addition, seven cheeses that were visibly mouldy, but not considered as either softripened or blue cheeses, were purchased. The products were reflective of a variety of origins, principally the USA, the United Kingdom (UK), and Europe. The details of each cheese, such as the country of origin, species of milk used in its preparation, and the species of fungus used for ripening, are presented in the Supplemental Data (Table S1; available from the corresponding author).
The samples were purchased between February and May 2021 and were either small individual cheeses or wedges from larger cheeses. Camembert and Brie are often sold as individual small round cheeses from which 200-500 g subsamples were taken. Blue-veined cheeses were generally purchased as wedges of larger "wheels" of cheese of 2-4 kg before being cut. Thus, samples were typically received as individually wrapped wedges cut from larger blocks (or wheels) by the grocer.
As a control, a minimally processed form of cheese, containing many of the bacteria typically present in cheese but lacking the fungal inoculum present in the soft-ripened or blue cheeses, was required. Cheese curds were selected for this purpose. The curds did not contain deliberately added fungi or the dye annatto. Curds were purchased from local (Peoria, IL, USA) groceries in vacuum-sealed bags and kept at 4°C until use.

Spiking of cheese
Chilled curds were grated, mixed, frozen with dry-ice and ground to the consistency of flour using a coffee mill as described previously (Maragos 2021). The powdered curds were kept at −20°C overnight, in a container with a loosely fitting lid to allow the dry ice to evaporate. It was essential to provide enough time for the CO 2 to fully evaporate before spiking. The powdered cheese was weighted in 20 g portions and spiked with CPA at six levels (15,30,75,150,300, and 750 µg/kg). An aliquot of spiking solution was added to the powdered cheese and mixed, and the solvent was evaporated for 15 min at ambient temperature. Details of the concentrations and the volumes of the spiking solutions used are listed in Table S2 (available from the corresponding author). Five replicate samples were prepared at each of six spiking levels (30 spiked samples total).

Sample preparation and extraction
The distribution of fungi within the cheese was nonhomogeneous. With soft-ripened cheeses, the fungi were visible in the rind, whereas for blue cheeses there were "veins" of fungal growth visible throughout the cheese. To improve homogeneity samples were grated first, where possible. For soft cheeses, the samples were cubed and mixed by kneading. This was needed to ensure a uniform distribution of the rind. Twenty g sub-samples were placed into 50 mL polypropylene centrifuge tubes and stored at −20°C until extraction. On the day of the analysis, samples were thawed and mixed with 100 mL of MeOH/H 2 O (4 + 1, v/v) in a 350 mL Pyrex® beaker. The mixture was blended for 90 s using an immersion blender. The mixture, containing 0.2 g equivalent cheese/mL of solution, was filtered (Whatman 2 V, Whatman plc, Maidstone, UK). The extract was diluted 1 + 9 (v/v) with PBS. Control cheese curds were similarly extracted, and the diluted extracts were used for the preparation of matrix matched calibration curves. For dilutions above 1:10, the diluent used was control extract (at 0.02 g equiv./mL). In this fashion, it was possible to ensure that the calibrants contained similar amounts of cheese matrix and MeOH as the samples. Diluted extracts were tested within 8 h of preparation. CPA calibrants were prepared over the range of 0.1 to 20 ng/mL. Fresh calibration standards were prepared daily.

ELISA procedure
Polystyrene microwell plates were coated with CPA-BLG at 1 μg/mL and blocked with PVA as described previously (Maragos et al. 2017). Plates with immobilised antigen were washed twice with 0.32 mL Tween-PBS. Test solutions were prepared in a polypropylene plate (Corning Inc., Corning, NY, USA) by mixing equal volumes of CPA mAb and test sample. The CPA mAb #1418 was prepared as a 1:30,000 dilution (approximately 70 ng/mL) in 1% (w/v) BSA-PBS. The wells of the polystyrene plate, with CPA-BLG immobilised, were washed twice with Tween-PBS and 0.1 mL of test solution was transferred into each well. After incubation at ambient temperature for 30 min the plate was washed three times and 0.1 mL of goat anti-mouse peroxidase conjugate (diluted 1:2,000 in BSA-PBS) was added. The plate was incubated for 30 min at ambient temperature then washed four times before addition of 0.1 mL of o-phenylenediamine substrate, prepared as described previously. Substrate was incubated at ambient temperature for 5 min, and the reaction was stopped by adding 0.1 mL of 1 N hydrochloric acid. Colour development was determined with a Synergy Neo microplate reader (Bio-Tek, Winooski, VT, USA) at 490 nm, with background correction at 650 nm.

Isolation of fungi and fungal cultures
Fungi were isolated from eight cheese samples: one Camembert and seven blue-veined cheeses. The blueveined cheese samples were selected based on the results of the CPA ELISA. One was a blue cheese selected because it contained a high level of CPA. Three others were Gorgonzolas selected because they contained high, medium, or no detectable CPA. Two others were Stiltons, one with a high CPA level and the other with no detectable CPA. The final sample was a Mitibleu with no detectable CPA.
To isolate fungal strains from cheese samples, a modification of the procedure described by Banjara et al. (2015) was used.
Briefly, 5 g of cheese was combined with 25 mL sterile 1% (w/v) peptone in a 50 mL conical tube and then vortexed at maximum speed until the cheese was uniformly suspended. In some cases, it was necessary to break up the cheese with a spatula prior to or in between vortexing. The resulting solution was then serially diluted with 10-fold dilutions to 1 × 10 −5 . One hundred µL of each dilution was plated onto YGC agar medium (0.5% yeast extract, 2% glucose, 1% agar (w/v)) that was amended with 0.1% (w/v) chloramphenicol and contained in a 100-mm Petri plate. The plates were then incubated aerobically in the dark at 25°C for 5 days. Selected colonies of filamentous fungi were transferred to a fresh YGC agar medium with 1% chloramphenicol and then incubated in the dark at 25°C for 4 days (Banjara et al. 2015). The resulting cultures were stored at 4°C until they were analysed for CPA production.
For analysis of CPA production, fungal strains were grown in two media: 1) cracked corn kernel medium prepared by soaking 2.5 g cracked kernels in 2 mL distilled water for 30 minutes and 2) a liquid medium (CZB-TE) prepared by mixing 1 L Czapek's broth medium with 0.2 mL trace elements solution (9.0 mM boric acid, 0.26 M citric acid, 0.19 M ZnSO 4 -6 H 2 O, 25 mM Fe(NH 4 ) 2 (SO 4 )-6 H 2 O, 10 mM CuSO 4 -5 H 2 O, 3.3 mM MnSO 4 , 2.0 mM Na 2 MoO 4 -2 H 2 O) (Leslie and Summerell 2006). Both media were autoclaved for 20 min at 120°C. Five hundred µL of a suspension of approximately 1 × 10 5 conidia per mL water was added to the cracked kernel medium in an 8-dram glass vial and 50 mL of CZB-TE in a 250-mL Erlenmeyer flask. The resulting cultures were incubated in the dark at 25°C for 14 days without shaking. In addition to fungal strains isolated from cheeses, we examined CPA production in three reference strains of P. camemberti from the U.S. Department of Agriculture, Agricultural Research Service (ARS) Culture Collection (NRRL, Peoria, IL, USA): strains NRRL 874, NRRL 875, and NRRL 885.

Extraction and ELISA of fungal cultures
One of the two extractions was performed depending upon the type of culture (cracked corn or CZB-TE). Cultures grown on cracked corn (3.5-4 g) were extracted by shaking with 25 mL of 0.1% formic acid in acetonitrile for 1 h at ambient temperature. The extract was filtered through a Whatman 2 V paper and then diluted 1 + 1 (v/v) with acetonitrile and filtered through a 13 mm 0.2 µm PVDF syringe filter and tested without further purification. The liquid (CZB-TE) cultures were extracted by adding 30 mL of 0.1% (v/v) formic acid in acetonitrile to 10 mL of culture. The mixture was shaken for 15 min at ambient temperature. For the liquid cultures, a phase separation was observed. A portion of the acetonitrile phase was removed and diluted with PBS for measurement of CPA by ELISA. ELISAs were conducted in the same manner as for the cheese samples. For initial screening, extracts were diluted from 1:40 to 1:100 with BSA-PBS. Depending upon the results from the initial tests, samples were re-tested at dilutions of up to 1:50,000 to quantify the CPA.

Data analysis
To permit comparisons across multiple days, raw absorbance data were transformed by dividing the absorbance of the sample or standard (B) by the absorbance of the toxin-free control (Bo). The transformed data, as percentage (B/Bo) were fit using a 4-parameter logistic dose-response equation (equation #8013, TableCurve, Systat Software, Inc., San Jose, CA, USA). For recovery studies, each replicate was tested with a minimum number of triplicate wells on a plate. Individual calibration curves were prepared for each microplate and the concentrations of toxins in the samples were determined by comparing the transformed data from samples to standard curves prepared on the same plate. For recovery studies, a total of 10 plates were used, each with its own calibration curve. Data obtained by averaging the 10 individual curves were used to calculate the limit of detection (LOD) and the concentrations resulting in 20% inhibition (IC 20 ), 25% inhibition (IC 25 ), 50% inhibition (IC 50 ) and 80% inhibition (IC 80 ). The LOD was calculated as the concentration required to give an absorbance of 3 standard deviations from the control (toxin-free) sample.

ELISA parameters
Calibration curves were obtained using CPA standards prepared in 1:10 dilutions of control matrix, by averaging results from 10 microplates (Figure 2). For ELISAs, sensitivity (LOD) is frequently calculated from the standard deviation of analyte-free samples, as the concentration of analyte required to give signal 3 standard deviations from the analyte-free controls. The LOD was 0.167 ng/mL. Assuming a 1:10 dilution of sample extract (i.e. 0.02 g equivalent of cheese/mL extract), this LOD was equivalent to 8.4 µg/kg in cheese. The limit of quantification (LOQ) for ELISAs is often calculated as the concentration required to observe a 20% change in signal (IC 20 ). Here, the IC 20 was 0.131 ng/mL, equivalent to 6.6 µg/kg in cheese. As the calculated LOQ was less than the LOD, we chose to use the IC 25 as the LOQ. As such, the LOQ was 0.168 ng/mL, equivalent to 8.4 µg/kg in cheese. The midpoint of the calibration curve (IC 50 ) was 0.451 ± 0.07 ng/mL (22.5 ± 3.5 µg/kg), and the upper end of the dynamic range, the IC 80 , was 1.658 ng/mL (83 µg/kg).

Recovery from spiked cheese curds
Control cheese curds were amended with CPA at levels from 15 to 750 µg/kg. Five replicate samples were spiked at each level (n = 30 total). Recoveries ranged from a low of 85 ± 25% at 300 µg/kg to 112 ± 46% at 15 µg/kg (Table 1). Overall recovery, from all samples, was 102.5 ± 26.8%.

Figure 2.
Calibration curve of CPA in cheese matrix containing 8% methanol. Points represent the averages from 10 plates ±1 standard deviation.

Survey of cheeses marketed in the USA
The edible portion, including the rinds, of a total of 254 samples of cheeses were tested by ELISA for CPA. Results for individual samples are presented in Table  S1. Of these, 79 were soft-ripened cheeses (Brie, Camembert), 168 were blue-veined cheeses, and 7 were miscellaneous cheeses that were also ripened with fungi.
Overall, CPA was found in 79 of the 254 samples (31.1%). To examine the differences between the softripened cheeses, which were ripened with P. camemberti and the blue-cheeses (ripened with P. roqueforti), the data were parsed by the type of cheese ( Table 2). The Brie group had the greatest percentage of CPA-positive samples (60.7%) and the DanaBlu group the lowest (0%). The highest median concentration in the positive samples was seen with the Siltons (447 µg/kg), but it should be noted that this represented only two positive samples, of which one was 870 µg/kg (

CPA in cultures obtained from cheeses and from reference strains of P. camemberti
All of the filamentous fungi recovered from cheeses and analysed for CPA production displayed characteristics consistent with the morphology of P. camemberti and P. roqueforti. That is, colonies on YGC agar medium were white to grey-green with dense mycelia and abundant round to ovoid conidia. Twenty of the fungal isolates recovered from cheese were analysed for production of CPA, including eight isolates from cheese samples that contained high levels of CPA. These included a Valdeon (C-115) with 1,250 µg/kg of CPA and three Gorgonzola samples with CPA levels ranging from non-detectable to 834 µg/kg (C-132, C-151 and C-149). Despite the high levels of CPA in some cheese samples none of the fungal isolates produced significant levels of CPA in either solid culture (cracked corn) or liquid culture (CZB-TE). The LOQ for CPA was 13 ng/ mL in liquid culture and 59 µg/kg in cracked corn. In contrast, the reference strains of P. camemberti produced 14.9-82.6 µg/mL in liquid culture.

Discussion
CPA has previously been found at low levels in animal milks. Dorner et al. (1994) conducted a transmission study by administering CPA orally to ewes at a dose of 5 mg/kg bw for 2 days. The average CPA concentration found in the milk, 24 h after the first dose, was 236 µg/kg. The highest level found was 568 µg/kg. Within 9 days, the CPA was no longer detected. Very few surveys have been done for CPA in milk. Of these, Losito et al. (2002) found 3 of 20 samples of pasteurised bovine milk contained CPA at 4.5, 6.2, and 8.3 µg/L. In two studies, Oliveira et al. found CPA in 2 of 48 samples (Oliveira et al. 2006) and 3 of 50 samples (Oliveira et al. 2008) of milk from Brazil. The five positive samples ranged from 6.4 to 9.7 µg/L. Given that P. camemberti can produce CPA and that this organism is used to ripen cheeses, more studies have focused on CPA in this matrix. CPA was found in 11 of 20 samples of Camembert cheese, predominantly in the rind, with 3 samples having 0.05 to 0.1 mg/kg, 5 samples having 0.1 to 0.2 mg/kg, and 3 samples having 0.4 to 1.5 mg/kg (Le Bars 1979).
Taleggio is a naturally ripened cheese from Italy. Of the six samples, five contained CPA, with the highest level (250 µg/kg) found in the rind of one sample (Finoli et al. 1999). There was also a correlation between the level of CPA and the amount of surface mould. A solid phase micro-extraction-LC method based upon UV detection (SPME-HPLC) was developed to measure CPA in Italian "white surface cheeses" (Zambonin et al. 2001). Of the six cheeses examined, CPA was found at levels from 20 to 80 µg/kg. In a later publication, a similar method was applied to 20 samples of Italian ewe's cheese. While the levels in the samples were not fully described, it was noted that some of the samples contained 0.4 to 2 mg/kg CPA (Monaci et al. 2007). An LC-MS/MS method using stable-isotope dilution was reported with a limit of detection of 0.
2 µg/ kg in cheese (Ansari and Häubl 2016). The method involved detection of CPA in positive mode MS/MS in 26 cheese samples. The samples were split into two categories, those which were Camembert, Brie, and Saint Albray varieties (group 1) and "other soft surfaceripened cheeses" (group 2). Levels of CPA in the first group ranged from 2.52 to 309 µg/kg, while those in the second group ranged from 0 to 3,610 µg/kg. Notably, 13 of the 14 samples in the second group were below 25 µg/kg, with one sample being exceptionally high (3,610 µg/kg). CPA was only detected in the outer layer (e.g. rind, 5 mm thickness) and not in the interior of the cheeses. Recently, an imaging surface plasmon resonance method and an HPLC-UV reference method did not detect CPA in seven samples of Camembert cheese, above a threshold of 40 µg/kg (Hossain et al. 2019). Penicillium camemberti does not produce CPA under all conditions. A study by Cisarova et al. (2012) of P. camemberti cultured on Czapek yeast extract agar (CZA) indicated that CPA production was related to temperature. Of 14 strains tested, two failed to produce CPA over the temperature range examined (0°C to 25°C). Based upon the literature, CPA might be expected to be a common contaminant of soft-ripened cheeses, with most samples containing less than 500 µg/kg and with occasional samples containing higher levels. Our results with the soft-ripened cheeses marketed in the United States fit this pattern. Forty-six per cent (36/79) of the soft-ripened cheeses contained detectable CPA, with the positive samples containing 12 to 3,820 µg/kg (median, 48.5 µg/kg, Table 2). Interestingly, samples of Brie were more likely to contain CPA and to contain higher levels, than samples of Camembert. The reason for the difference is unclear. Although Brie and Camembert have different flavours and aromas, the two are similar in production. Both are ripened with P. camemberti. One major manufacturer of soft-ripened cheeses has indicated that, while similar techniques are used in production, the lactic acid bacteria used as starters are "stronger" in Camembert than in Brie (Président Cheese 2022). We speculate that perhaps the use of the milder bacterial strains provides an environment more conducive to the production of CPA by P. camemberti.
Cambozolas are a type of cheese where two fungi are used for ripening: the surface is ripened with P. camemberti, while the interior is ripened with P. roqueforti. We detected CPA in 6 of the 12 Cambozola's tested. These contained a median level of 27 µg/kg CPA, a level similar to that of the Bries and the Camemberts (Table 2). Of surprise to us was the presence of CPA in other varieties of blue-veined cheeses. This was a surprise because we did not expect P. roqueforti, the major fungus used to ripen "blue" cheeses, to produce CPA. Excluding the Cambozolas, 35 of 156 samples (22.4%) of the blue-veined cheeses contained detectable CPA. The significant percentage of samples that contained CPA, where P. camemberti, does not appear to have been intentionally used in the ripening process, was intriguing.
Logically, the CPA present in such samples could arise from (at least) three sources. The first of these could be the carryover of CPA from contaminated milk into the cheese. The aforementioned studies suggest this as a possibility, although we are unaware of any studies to examine the fate of CPA during cheese production. The second potential source would be through production by the fungi used in the ripening. For the blue cheeses that fungus, P. roqueforti, is not known to produce CPA. The third potential source would be the introduction of other fungi, such as P. camemberti, before, during, or after the cheese-making process. This implies the potential for P. camemberti to be among the fungi used in ripening of blue cheeses, perhaps as a contaminant of the milk or the culture used for ripening. Since cheeses are often handled by multiple parties before reaching the consumer, the potential for contamination of blue cheeses with P. camemberti during distribution (cutting, repackaging, shipment) is a distinct possibility. Although the route of introduction is unknown, the recent finding of P. camemberti in a sample of blue cheese suggests that this organism may be present in such cheeses (Ribeiro et al. 2020).
To investigate the possibility that P. roqueforti, rather than P. camemberti might be producing CPA, samples of several blue cheeses with various levels of CPA contamination were cultured. The predominant isolates from each cheese were tested for their ability to produce CPA on cracked corn and in liquid (CZB-TE) medium. None of the isolates from blue cheeses produced CPA under either culture condition.
An isolate from a Camembert cheese that lacked CPA also did not produce CPA. The latter is not surprising, as while many strains of P. camemberti produce CPA, such production is not assured (Karlshøj et al. 2007). To determine whether the culture conditions were causing the lack of production, three reference strains of P. camemberti from the ARS Culture Collection were analysed for production on CZB-TE. All produced high concentrations of CPA (14-83 µg/mL), suggesting that the culturing conditions were conducive to CPA production. While the high prevalence of CPA in blue cheeses, albeit at low concentrations, is intriguing, the source of the CPA remains unresolved. The gene cluster that confers production of CPA has been identified in Aspergillus species (Chang et al. 2009;Tokuoka et al. 2015). The National Center for Biotechnology Information's GenBank database currently includes genome sequence data for six strains of P. camemberti and five strains of P. roqueforti. BLASTn analysis of these genome sequences using DNA sequences of the Aspergillus CPA biosynthetic genes revealed that a homolog of the CPA cluster is present in all P. camemberti strains but none of the P. roqueforti strains. This finding is consistent with reports of production of CPA by P. camemberti but not by P. roqueforti.

Conclusions
CPA was detected in 36/79 (45.6%) of the soft-ripened cheeses and, somewhat surprisingly, in 41/168 (24.4%) of the blue-veined cheeses. The maximum levels found were 3,820 µg/kg in a sample of Brie, 1,250 µg/kg in a sample of blue, and 7,900 µg/kg in a sample of Monte Enebro. Median levels in positive samples were much lower than this: 48.5 µg/kg and 30 µg/kg in the softripened and blue cheeses, respectively. There is no consensus value for the acceptable daily intake of CPA. Proposed values have ranged from 0.1 to 10 µg/kg bw/ d. For a 70-kg adult, this would equate to a CPA intake of 7 to 700 µg/day. The majority of the soft-ripened cheeses (54.4%) and blue cheeses (75.6%) tested herein did not contain detectable CPA, while median levels in positive samples were relatively modest (30 to 48.5 µg/kg cheese). This suggests that it is unlikely that individuals will attain an intake of 700 µg CPA/d through eating cheese. However, 10.1% of the soft-ripened cheeses contained more than 700 µg/kg of CPA. For blue cheeses 2.4% exceeded this level. Consumption of 10 g of such a cheese would exceed the more stringent (7 µg) threshold, while it would require consumption of 1 kg to exceed the higher (700 µg/kg) threshold. Ultimately, determining the significance of the levels found in cheeses will require the establishment of a consensus acceptable daily intake and improved consumption data.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
This work was supported by the U.S. Department of Agriculture, Agricultural Research Service, Projects 5010-42000-052-00D and 5010-42000-053-00D. Disclaimer Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the U.S. Department of Agriculture. USDA is an equal opportunity provider and employer.