Quantitative PCR reveals the frequency and distribution of 3 indigenous yeast species across a range of specialty cheeses

Indigenous microorganisms are important compo-nents of the complex ecosystem of many dairy foods including cheeses, and they are potential contributors to the development of a specific cheese’s sensory properties. Among these indigenous microorganisms are the yeasts Cyberlindnera , , and Kazachstania servazzii , which were previously detected using traditional microbiological methods in both raw milk and some artisanal specialty cheeses produced in the province of Québec, Canada. However, their levels across different cheese varieties are unknown. A highly specific and sensitive real-time quantitative PCR assay was developed to quantitate these yeast species in a variety of specialty cheeses (bloomy-rind, washed-rind, and natural-rind cheeses from raw, thermized, and pasteurized milks). The specificity of the quantitative PCR assay was validated, and it showed no cross-am-plification with 11 other fungal microorganisms usually found in bloomy-rind and washed-rind cheeses. Cyberlindnera jadinii and P. kudriavzevii were found in the majority of the cheeses analyzed (25 of 29 and 24 of 29 cheeses, respectively) in concentrations up to 10 4 to 10 8 gene copies/g in the cheese cores, which are considered oxygen-poor environments, and 10 1 to 10 4 gene copies/ cm 2 in the rind. However, their high abundance was not observed in the same samples. Whereas C. jadinii was present and dominant in all core and rind samples, P. kudriavzevii was mostly present in cheese cores. In contrast, K. servazzii was present in the rinds of only 2 cheeses, in concentrations ranging from 10 1 to 10 3 gene copies/cm 2 , and in 1 cheese core at 10 5 gene copies/g. Thus, in the ecosystems of specialty cheeses, indigenous yeasts are highly frequent but variable, with certain species selectively present in specific varieties. These results shed light on some indigenous yeasts that establish during the ripening of specialty cheeses.


INTRODUCTION
A consistent cheese quality and flavor can be difficult to achieve from batch to batch, potentially due to the complex microbial ecosystem and the presence of indigenous microorganisms (Beresford and Williams, 2004). Recent studies have demonstrated that the cheese ecosystem is composed of hundreds of adventitious microbial species that emerge during the ripening process (Bokulich and Mills, 2013;Wolfe et al., 2014). They may be introduced into milk at the farm or the dairy plant (Bokulich and Mills, 2013).
Yeasts are natural members of the cheese ecosystem because of their capacity to grow in wide ranges of pH and temperature and their tolerance to high salt concentrations and low water activity (Fleet, 2011). Because of their common presence, multiple techniques are used to detect, identify, and quantify yeasts in cheese, including traditional microbiology techniques, quantitative (q)PCR, and omics methods (Lessard et al., 2014;Wolfe et al., 2014;Dugat-Bony et al., 2016).
To date, few studies have addressed the frequency and distribution of indigenous yeasts in cheeses, despite their potential effects on the physicochemical composition of the cheeses and their microbial ecosystem (Wyder and Puhan, 1999;Chen et al., 2012).
In raw milk or artisanal specialty cheeses from the province of Québec, P. kudriavzevii was isolated and identified, using ITS1-5.8S-ITS2 rDNA sequencing, in numerous cheese samples, whereas C. jadinii and K. servazzii were found less frequently (Lavoie et al., 2012). Because the approach was not quantitative, the abundance of these yeasts across different cheeses and their ability to influence the cheese ripening process remain unclear.
The aim of this study was to develop a qPCR method to determine the concentrations of C. jadinii, P. kudriavzevii, and K. servazzii in different areas (core and rind) of several types of commercial specialty cheeses.

Biological Material and Genomic DNA Extraction
Reference fungal strains from the Laboratoire de Mycologie Alimentaire (LMA, Université Laval, Québec, Canada) collection were used to develop a qPCR assay (Table 1). Strains were cultured from culture stock stored at −80°C in YEG medium (10 g/L yeast extract, Bio Basic; 10 g/L glucose, EMD Chemicals) containing 15% glycerol, and were plated on YEG agar (15 g/L bacto agar; BD Diagnostics). Individual colonies were cultured in YEG liquid medium and incubated at 25°C for 48 h. Yeasts were pelleted by centrifugation (5 min at 20,000 × g at room temperature), which were used for genomic DNA extraction using a phenol-chloroform extraction protocol (Al-Samarrai and Schmid, 2000). Yield and quality of DNA were determined using a Nanodrop spectrophotometer (ND-1000, ThermoFisher Scientific).

Cheese Samples and Total DNA Extraction
Triplicates of each of 29 ready-to-eat specialty cheese varieties from 22 cheese plants (for a total of 87 individual cheeses sampled) from the province of Québec, Canada, were sampled. They were classified according to their variety: washed rind (n = 15), bloomy rind (n = 12) or natural rind (n = 2), and by heat treatment of the milk: raw (n = 6), thermized (n = 14), or pasteurized (n = 9), and were coded for confidentiality (Table 2). Thermization was conducted by each cheesemaker according to the regulation of the province of Québec; that is, between 15 and 20 s at 57 to 68°C. The cheeses were stored at 4°C for 3 d. A total of 50 cm 2 was sampled and weighed for each cheese rind; in brief, 25 cm 2 was scraped from each flat surface to recover the microbiota and a minimum of the cheese matrix underneath (<1 mm). In addition, 10 g of each cheese variety's core was sampled for 1 of 3 specimens for each of the 29 cheeses. All rind and core samples were stored at −80°C before grinding using a liquid nitrogen-cooled CryoMill (Retsch). Grinding consisted of a precooling cycle at 5 Hz, followed by a 4-min grinding step at 25 Hz.
Cheese total DNA extraction was performed in triplicate for each cheese sample using a Purelink viral RNA/DNA kit (Invitrogen/Life Technologies) with the following modifications: a total of 30 to 40 mg of each ground cheese sample was homogenized in 200 µL of LMA-1019 (commercial cheese starter) Lessard et al., 2012 Geotrichum candidum LMA-1028 (commercial cheese starter) Lessard et al., 2014 0.9% NaCl, before the addition of proteinase K and lysis buffer from the kit. After DNA elution, RNA was removed using a 0.5 mg/mL RNase A treatment for 1 h at 37°C. Yield and quality of the extracted DNA were measured using a Synergy HI microplate reader and Gen5 software (BioTek).

qPCR Absolute Quantification Using Standard Curve Assays
The qPCR target genes (CEM1, GSTY-1, and CenH3) were individually PCR amplified, cloned in the pCR-II TOPO vector, and transformed in Escherichia coli TOP10 using a TOPO TA Cloning kit (Invitrogen/ Life Technologies). Recombinant plasmid DNA was purified with a Plasmid DNA Maxiprep kit (Qiagen), and then sequenced. To use the recombinant plasmid DNA as a quantification standard, the concentration was measured with a Synergy HI microplate reader and Gen5 software (BioTek). The number of gene copies was calculated using the following formula: Thermization was conducted by the cheesemaker according to the regulation of the province of Québec; that is, between 15 and 20 s at 57 to 68°C.

2
No raw or pasteurized natural-rind cheeses were available at the time of sampling. When used as a quantification standard during qPCR assays, the recombinant plasmid DNA was diluted 6to 10-fold and deposited in triplicate in 96-well plates. These serial dilutions allowed us to determine a limit of detection of 3 gene copies per PCR reaction. Standard curves were generated to determine the qPCR efficiency and dynamic range of each reaction (dynamic range for target gene CME1: 3 to 3 × 10 8 gene copies per reaction; for GSTY-1 and CenH3: 3 to 3 × 10 9 gene copies per reaction).
Quantitative PCR was performed on an ABI7500 Fast system apparatus (Applied Biosystems/Thermo-Fisher Scientific). The qPCR plates were set up using an epMotion 5075 VAC automated pipetting system (Eppendorf). Optimal dilutions of DNA samples were chosen to fit the dynamic range of the standard curves and to ensure the absence of PCR inhibitors. The qPCR reagents were used as described by Lessard et al. (2012). The qPCR program began with an initial denaturation step at 95°C for 20 s, followed by 40 cycles of denaturation 95°C for 3 s and annealing/extension at 60°C for 30 s. The qPCR data were analyzed using 7500 software v2.0.6 (ThermoFisher Scientific).

Statistical Analysis
A one-way ANOVA was used to analyze the effect of the type of cheese on the detection of each indigenous yeast being tested. Because of the non-normality of the data, P-values were adjusted using a permutation distribution. Variety of cheese was considered a fixed factor, and comparisons were made between specific types of cheese (rind type and heat treatment). SAS/ STAT software (2015; SAS Institute Inc.) was used for analysis; specifically, the GLM procedure was followed by the MULTTEST procedure.

qPCR Assay Efficiency
Highly specific and sensitive qPCR assays were developed for the quantification of 3 indigenous yeast species (C. jadinii, P. kudriavzevii, and K. servazzii) in samples originating from both the rind and core of specialty cheeses. The specificity of qPCR amplification was confirmed by sequencing the obtained qPCR amplicons and by negative qPCR detection using 10 other fungal species isolated from dairy products. BLAST searches of the CME1, GSTY-1, and CenH3 target genes revealed that these genes were present in only one copy in the genomes of C. jadinii [whole-genome sequencing (WGS) project BAEL01], P. kudriavzevii (WGS project JQFK01), and K. servazzii (unpublished WGS project LMA-647, Laboratoire de Mycologie Alimentaire), respectively. Resulting average qPCR efficiencies were 98.4, 97.8, and 96.6% for C. jadinii, P. kudriavzevii, and K. servazzii, respectively (R 2 values ≥0.990; Table 3).

Distribution of C. jadinii in Quebec's Specialty Cheeses
Cyberlindnera jadinii was detected in 25 of 29 cheese cores at concentrations of 10 5 to 10 8 gene copies/g (Table 4), and was detected in the rinds of 24 of 29 cheeses at 10° to 10 4 gene copies/cm 2 in all cheese varieties. It was absent in both the core and rind of only one cheese, C1, a washed-rind cheese made with raw milk. The population of C. jadinii in the cheese core was not dependent on the cheese variety (P > 0.05) but was correlated with the milk heat treatment applied (P < 0.01, Figure 1), especially in bloomy-rind cheeses made from raw and thermized milk (P < 0.001). At the cheese surface, high C. jadinii counts were obtained in bloomy-rind cheeses (12/29 cheeses) compared with washed-rind cheeses (15/29 cheeses; P ≤ 0.001, Table  4), although this difference was not observed in core samples taken from the same cheese varieties. Cheeses made from raw and thermized milk had more C. jadinii on their surface than pasteurized milk cheeses (P < 0.001), but we detected no differences between raw and thermized milk cheeses (Figure 2).
The presence of C. jadinii in multiple cheeses manufactured in the same cheese plant was verified by analyzing 2 different cheese varieties from each of 8 artisanal cheese plants (plant A: C1 and C4; plant B: C3 and C17; plant C: C5 and C19; plant D: C6 and C8; plant E: C9 and C10; plant F: C12 and C27; plant G: C16 and C18; plant H: C20 and C29, Table  4). Counts of C. jadinii on the surface of a bloomy-rind and a washed-rind cheese (C17 and C3) produced at the same cheese plant were significantly different (10 4 gene copies/cm 2 and 1 gene copy/cm 2 , respectively; P < 0.001). The same significant difference in C. jadinii counts was observed in the cores of these cheeses: C. jadinii was more abundant in bloomy-rind cheeses than in washed-rind cheeses (P < 0.001; Table 4).

Distribution of P. kudriavzevii in Quebec's Specialty Cheeses
Pichia kudriavzevii was detected on the surface of 6 varieties of cheese at 0.9 ± 1.5 to 1.2 ± 2.1 × 10 2 gene copies/cm 2 , and in the core of 24 cheeses from 2.4 ± 4.2 × 10 5 to 4.15 ± 1.03 × 10 7 gene copies/g (Table  5). It was not detected in the rind or core of 4 cheeses (C1, C12, C15, and C20). The type of heat treatment applied to the milk used for each cheese did not affect the presence or levels of P. kudriavzevii in the cores of washed-rind cheeses. In the bloomy-rind cheeses, however, a significantly higher count of P. kudriavzevii was detected in raw milk cheeses than in thermized and pasteurized milk cheeses (P < 0.05; Figure 3). This indigenous yeast was detected in the core of 11 of 12 bloomy-rind cheeses in amounts ranging from 10 5 to 10 7 gene copies/g. Finally, different cheeses produced in the same plant showed different levels of P. kudriavzevii. For example, a difference was observed between washed-rind and bloomy-rind cheeses C3 and C17 from cheese plant B (10 5 and 10 7 gene copies/g, respectively).
Only 1 of the 2 cheese varieties from cheese plant H, which produced the bloomy-rind cheese C20 (< limit of detection) and the natural-rind cheese C29 (10 7 gene copies/g), contained P. kudriavzevii.
jadinii was rarely reported in cheese. It had been isolated from, but not quantified in, the rind of a Livarot cheese (a soft, red washed-rind cheese made from raw milk), during an early stage of ripening before salting (Larpin-Laborde et al., 2011). It was also reported in the core of a Portuguese Serpa cheese (Gonçalves Dos Santos et al., 2017). In an earlier study by our group, using traditional microbiological methods, C. jadinii was isolated in only a single raw milk sample and a single semi-hard washedrind cheese core made with thermized milk (Lavoie et al., 2012). Interestingly, the qPCR method detected and quantified C. jadinii in 83% of surfaces and 86% of cores, a higher frequency than previously reported. The low detection rate in previous studies could be explained by the lack of specific detection methods (Cogan et al., 2014) or by the presence of C. jadinii as a subdominant species of the cheese ecosystem because it is masked by other microorganisms. Given that C. jadinii was detected in relatively large quantities in several cheese varieties, the effect of C. jadinii on the physicochemical characteristics and microbial ecosystem of cheeses during ripening warrants further investigation.
Kazachstania servazzii was detected in only 3 varieties of specialty cheese analyzed in this study: on 2 natural-rind cheeses (C28 and C29) and in 1 washed-rind cheese core (C12). Kazachstania servazzii was found in the rind (surface) of all 3 biological and 27 technical replicates of C28. It was also isolated from the same cheese in 2012 (Lavoie et al., 2012), suggesting that this yeast is well established in the cheese plant where C28 is produced. Kazachstania servazzii is rarely reported in cheese, with the exception of a Camembert-type cheese produced using a Tibetan freeze-dried kefir coculture as a starter (Mei et al., 2014). In that particular case, this yeast was predominant during the beginning of the ripening process (d 0 to 15; Mei et al., 2014): 51 isolates of this species were found, compared with 15 isolates of Debaryomyces hansenii (Mei et al., 2014). This ecosystem clearly evolved during the ripening process, as D. hansenii was found in higher concentrations at the end of the ripening period (d 10 to 35) of the cheese. These results, taken in conjunction with the current report, suggest that K. servazzii is not well adapted to survive in the cheese matrix or that it is displaced by other yeasts during ripening. This could be explained by the ability of K. servazzii to metabolize galactose but not lactate (Vaughan-Martini et al., 2011). Depending on techniques used during ripening (e.g., lactic vs. rennet coagulum, soft vs. hard cheese), residual galactose can be limited relative to lactate during the ripening process (Fox and McSweeney, 2004).
In previous work, P. kudriavzevii was found in raw milk from 8 farms and was the most frequently isolated yeast species from ripened cheese: 7 isolates came from cheese rinds and 12 from cheese cores (Lavoie et al., 2012). Pichia kudriavzevii is naturally present in milk, can survive the cheesemaking process, and can be found in cheese. In this study, P. kudriavzevii was found in 26 of 29 cheeses analyzed (6 surface and 24 cores), in all cheese rind varieties (washed, bloomy, and natural rind) and in cheese from milk submitted to all types of heat treatment (raw, thermized, and pasteurized). Our results confirm that P. kudriavzevii is frequently found in cheeses; it has been reported in Austrian fresh and sour curd cheeses (Lopandic et al., 2006), Graukäse cheese (Prillinger et al., 1999), and in multiple raw milk cheeses (Borelli et al., 2006;Pangallo et al., 2014;Chombo-Morales et al., 2016). Pichia kudriavzevii was found to be one of the dominant species on the surface of Cotija raw milk cheese, as identified by PCR-denaturing gradient gel electrophoresis (PCR-DGGE;Chombo-Morales et al., 2016). This evidence, in conjunction with our results, suggests that the presence of P. kudriavzevii in cheese might have been underestimated in previous studies using traditional microbiological methods (Valdés-Stauber et al., 1997;Cogan et al., 2014).
Although P. kudriavzevii appears to be common in cheese microbial ecosystems, its contribution to the cheese ripening process remains unclear. It is generally accepted that yeasts are found in 10-to 1,000-foldhigher concentrations on the surface of cheese than in the core because of their oxygen and nutritional requirements (Schmidt et al., 1980;Beresford and Williams, 2004;Chamba and Irlinger, 2004). However, our qPCR quantification method showed that the indigenous yeasts C. jadinii and P. kudriavzevii can be found in high concentrations (up to 10 4 -10 8 gene copies/g) in the cheese core compared with the surface (10 1 -10 4 gene copies/cm 2 , equivalent to 10 2 -10 5 gene copies/g). As the presence and metabolism of indigenous yeasts in the core and surface could affect the flavor development of the cheese (Beresford and Williams, 2004), further investigations of the role of C. jadinii and P. kudriavzevii in cheese ripening are warranted. Furthermore, the presence of these particular yeasts may reflect intrinsic properties, such as the ability to grow under low oxy-gen conditions or in the presence of lactate and citrate (Kurtzman, 2011a,b).
Previous studies have shown that the type of heat treatment applied to milk affects the biodiversity of microorganisms found in cheese made from that milk (Delcenserie et al., 2014;Wolfe et al., 2014;Irlinger et al., 2015). The quantity of C. jadinii in the cheese   H) were produced by the same cheese plant. (Cheeses with letters not repeated in this table indicate that another cheese produced by this cheese plant was analyzed but the microorganism was absent.) Means and SE were calculated from means of each of 3 biological replicates (3 technical replicates were performed per biological replicate) for the rinds of each cheese variety, and from the means of values obtained from 3 technical replicates for each core (1 biological replicate was sampled per cheese variety). 2 LOD = limit of detection, determined using serial dilutions, and set to 3 gene copies per PCR reaction.  H) were produced by the same cheese plant. Means and SE were calculated from means of each of 3 biological replicates (3 technical replicates were performed per biological replicate) for the rinds of each cheese variety, and from the means of values obtained from 3 technical replicates for each core (1 biological replicate was sampled per cheese variety).
2 LOD = limit of detection, determined using serial dilutions, and set to 3 gene copies per PCR reaction.
core increased with milk thermization compared with cheese from raw and pasteurized milk (P < 0.01; Figure  1). Conversely, P. kudriavzevii was more abundant in bloomy-rind cheese made from raw milk (P < 0.05; Figure 3). However, caution should be exercised in the interpretation of these results because initial concentrations of yeasts in the milk were not established for the cheeses studied. Changes in indigenous yeast counts from the raw milk to the ripened cheese surface and core should be further studied to provide insight into the sources and effects of these microorganisms.

CONCLUSIONS
Using species-specific primers and probes, we developed a highly specific and sensitive real-time quantitative PCR method to analyze the presence and distribution of Cyberlindnera jadinii, Pichia kudriavzevii, and Kazachstania servazzii in specialty cheeses. Results confirmed that these 3 indigenous yeasts are present in different commercial cheeses produced in the province of Québec and revealed that C. jadinii and P. kudriavzevii are found in especially large quantities in the cheese core. This study highlights the importance of research not only on the microbial dynamics of the cheese core but also on the effects of these indigenous yeast species on the physicochemical properties of specialty cheeses.

ACKNOWLEDGMENTS
This work was supported by Novalait Inc., the Fonds de Recherche Nature et Technologies du Québec (FRQNT; Québec, Canada), the Québec Consortium for Industrial Bioprocess Research and Innovation (CRIBIQ; Québec, Canada), and the Natural Sciences and Engineering Research Council of Canada (NSERC; Ottawa, Canada; Grant ID 2016-LG-195927). The authors are grateful to Laval University's Service de Consultation Statistique for their assistance. The authors have not stated any conflicts of interest.