High Prevalence of Enterobacterales in the Smear of Surface-Ripened Cheese with Contribution to Organoleptic Properties

The smear of surface-ripened cheese harbors complex microbiota mainly composed of typical Gram-positive aerobic bacteria and yeast. Gram-negative bacteria are usually classified as un-wanted contaminants. In order to investigate the abundance and impact of Gram-negative bacte-ria naturally occurring in the smear of surface-ripened cheese, we performed a culture-based analysis of smear samples from 15 semi-hard surface-ripened cheese varieties. The quantity, di-versity and species distribution of Proteobacteria in the surface smear of the analyzed cheese vari-eties were unexpectedly high, and comprised a total of 22 different species. Proteus and Morganella predominated most of the analyzed cheese varieties, while Enterobacter, Citrobacter, Hafnia and Serratia were also found frequently. Further physiological characterization of Proteus isolates re-vealed strong proteolytic activity, and the analysis of volatiles in the smear cheese surface head-space suggested that Enterobacterales produce volatile organic flavor compounds that contribute to the organoleptic properties of surface-ripened cheese. Autochthonous members of Enterobac-terales were found in 12 of the 15 smear samples from surface-ripened cheeses, suggesting that they are part of the typical house microbiota that shape the organoleptic properties of the cheese rather than represent unwanted contaminants. However, further investigation on safety issues of the individual species should be performed in order to manage the health risk for consumers.


Introduction
The production of cheese has a long tradition in Europe that has given rise to a majority of commercially important cheese varieties [1]. About one third of the yearly cheese production of Switzerland comprises semi-hard cheeses, while the majority are surface-ripened varieties with washed rind such as Appenzeller or Raclette cheese [2].
The development of a complex microbial biocoenosis in the surface smear during the ripening process is characteristic of these semi-hard surface-ripened cheese varieties. This surface smear microbiota is not only responsible for the red-orange coloration of the cheese surface, it also greatly determines the organoleptic properties of this type of cheeseincluding its intense sulfuric smell [3]. When the equilibrium of surface smear microbiota is unbalanced, undesirable contaminants such as spoilage microorganisms or pathogens may grow on the cheese surface, which may cause rind defects or pose a potential threat to consumers' health [4].
The surface smear develops naturally on cheese surface that is regularly washed with brine and ripened at suitable conditions such as 13-15 • C and high relative humidity (>95%). At the end of the ripening process, the surface smear microbiota consists mainly

Cheese Samples
In total, 15 samples (V1-V15) of Swiss semi-hard surface-ripened cheeses were obtained from retail trade in Switzerland and subjected to microbial analysis of the cheese surface smear composition (Table 1). Cheese samples were acquired either as vacuum film-prepacked from self-service shelves or wrapped in cheese wrapper from the cheese counter. As indicated in Table 1, the cheese varieties included in this study were produced by different production systems and could be assigned thus: Mountain-like, Trappist-like, Raclette-like and Tilsit-like cheese. All the cheese varieties were produced from cow's milk, either raw, after thermization (light heat treatment step, usually between 57-68 • C for 15 s) or pasteurization (72-75 • C, 30 s), and varied in fat content in dry matter from 25-55%. weight of a whole round cheese wheel according to cheese manufacturer; b according to cheese manufacturer; c in dry matter according to cheese manufacturer.

Sampling Procedure and Microbial Analysis by Culturing
From all investigated cheeses, a rectangle of 8 cm 2 in size and 2-3 mm thickness (corresponding to 2.3 ± 0.2 g weight) was collected from the cheese surface and further processed as described in a previous study [13]. Colony-forming units (CFU) were determined by surface plating of serial dilutions of the homogenized cheese surface smear on Tryptic Glucose Yeast Agar (TGYA; Biolife, Milano, Italy), supplemented with 1% (w/v) peptone from casein (Merck, Darmstadt, Germany), Violet Red Bile Glucose Agar (VRBG; Biolife, Milano, Italy), Endo Agar (Endo; Biolife, Milano, Italy) and Plate Count Skim Milk Agar (PCSMA; Merck, Darmstadt, Germany), supplemented with 0.0005% (w/v) crystal violet (Roth, Karlsruhe, Germany) and 0.0005% (w/v) vancomycin hydrochloride (Roth, Karlsruhe, Germany) for inhibition of Gram-positive microbiota (PCAI) as described by Delbès et al., 2007 [10]. The incubation properties were as follows: total mesophilic aerobic bacterial counts on TGYA (3 days 30 • C aerobic incubation, followed by 7 days 22 • C room temperature incubation under daylight), Gram-negative colony counts on PCAI (2 days 30 • C aerobic for PCAI or anaerobic incubation for PCAI an ), Enterobacteriaceae colony counts on VRBG Agar (1 day 37 • C aerobic incubation) and coliform bacteria on Endo Agar (1 day 37 • C aerobic incubation) for isolation purposes. Colony counts on TGYA, PCAI and VRBG were determined as the weighted average.

Isolation of Gram-Negative Bacteria and Genotypic Identification by Partial 16S rRNA and gyrB Sequencing
Three representative colonies from each colony type identified on VRBG, PCAI, PCAI an or Endo plates were picked and purified by the streak plate method on the same culture media. Two additional subcultures were performed on half-concentrated Brain Heart Infusion Agar (BHI 1 2 ) containing 1.85% (w/v) BHI broth (Biolife, Milano, Italy), 0.25% (w/v) sodium chloride (Merck, Darmstadt, Germany) and 1.5% (w/v) agar (Roth, Karlsruhe, Germany).
For identification of bacterial isolates and comparative analysis by partial 16S rDNA or gyrB sequencing, DNA extraction was performed from 2 mL of cells grown overnight at 37 • C in BHI 1 2 containing 0.25% (w/v) sodium chloride. Initially, cells were washed twice in 1 mL phosphate-buffered saline (PBS) pH 7.4, containing 10 mM potassium chloride (Merck, Darmstadt, Germany), 137 mM sodium chloride (Merck, Darmstadt, Germany), 10 mM disodium hydrogen phosphate (Roth, Karlsruhe, Germany) and 2 mM potassium dihydrogen phosphate (Fluka Chemie GmbH, Buchs, Switzerland) [29]. DNA was extracted by the use of the GenElute TM Bacterial Genomic DNA Kit (Sigma-Aldrich Chemie GmbH, Steinheim, Germany) according to the manufacturer's instructions, with the exception of doubling heated incubation times and DNA elution performed with 50 µL preheated (55 • C) Milli-Q Water (Millipore AG, Zug, Switzerland). DNA concentration was determined using a ND-1000 Spectrophotometer (NanoDrop ® Technologies, Wilmington, USA).
Electropherograms of the resulting 16S rRNA or gyrB DNA sequences were carefully analyzed and manually edited when necessary. The sequences were compared to GenBank database sequences using the Basic Local Alignment Search Tool (BLAST) algorithm provided by the National Center for Biotechnology Information (NCBI; Bethesda, MD, USA) and SepsiTest TM BLAST online tool (Molzym GmbH & Co. KG, Bremen, Germany) [32]. Best BLAST hits of database sequence comparison were recorded. Percentage values of ≥99% identity of the query sequence to the closest relative database sequence were considered as species identification, whereas percentage values between <99% and ≥97% were considered as identification on a genus level. If the two different BLAST approaches from a single sequence resulted in different species assignments, or if the BLAST results from gyrB and 16S rDNA sequences revealed different species assignments, the valid genus designation was recorded.

Determination of Proteolytic and Lipolytic Activity of Gram-Negative Isolates
Selected Gram-negative bacteria were analyzed for their proteolytic and lipolytic properties by testing hydrolysis of proteins and fats embedded in solid media. Bacteria were grown overnight in 5 mL BHI 1 2 broth at 37 • C under aerobic conditions to a cell density of about 10 8 cells mL −1 , as verified by OD measurement on a spectrophotometer Libra S22 (Biochrom Ltd., Cambridge, England). For testing the proteolytic activity of isolates, 5 µL of overnight culture was spotted in triplicates on Caso Bouillon Skim Milk Agar (CBSMA) containing 4% (w/v) Caso Bouillon Broth (Roth, Karlsruhe, Germany), 5% (w/v) skim milk powder (Oxoid LTD., Basingstroke, Hampshire, England) and 1.5% (w/v) agar (Roth, Karlsruhe, Germany). Lipolytic activity was determined using 5 µL of an overnight culture spotted in triplicates on Tributyrin Agar (TBA) containing 0.5% (w/v) peptone from casein (Merck, Darmstadt, Germany), 0.3% (w/v) yeast extract (Merck, Darmstadt, Germany), 0.4% (w/v) tween 80 (Merck, Darmstadt, Germany), 1.5% (w/v) agar (Roth, Karlsruhe, Germany) and 1% (w/v) glycerintributyrate (Merck, Darmstadt, Germany). Incubation of each strain was performed under aerobic and anaerobic conditions at 4 • C, 22 • C (room temperature) and 30 • C for 12 days. Proteolytic and lipolytic bacterial activity was recorded as an agar clearing zone (halo) surrounding the bacterial growth after 2, 5 and 12 days of incubation. The extent of the proteolytic or lipolytic active zone was quantified by measuring the diameter of the halo in mm and subtracting the diameter in mm of the colony formed by the tested strain. The average for three tested colonies of the same bacterial strain and the according standard deviation were calculated. Clearing zones or overgrowing colonies with a diameter of more than 40 mm were recorded as a proteolytic or lipolytic activity of >40 mm.

Volatile Organic Flavor Compound Analysis of Cheese Surface Smear
A cheese portion of Tilsit-like variety V2, obtained directly from the manufacturer at the end of the ripening period of 70-110 days, was analyzed for volatile organic flavor compounds (VOC) released from the cheese surface smear by the Institute for Food Sciences (Agroscope, Berne, Switzerland). For this purpose, volatiles deriving from the cheese surface smear were identified by headspace solid-phase microextraction coupled with gas chromatography mass spectrometry (HS-SPME-GC-MS) or gas chromatography pulsed flame-photometric detection (HS-SPME-GC/PFPD). Individual chemical compounds detected in significant amounts in the VOC profiles were compared to volatile flavor compound patterns in bibliographic data to identify chemical compounds that presumably originate from the metabolic activity of members of Enterobacterales.

Abundance of Total Mesophilic Aerobic Bacteria, Gram-Negative Bacteria and Enterobacterales in Cheese Surface Smear
The weighted average colony counts in CFU cm −2 determined by culturing of the surface smear microbiota of 15 different semi-hard surface-ripened cheese varieties are illustrated in Figure 1. Counts for total mesophilic aerobic bacteria were in the range of 6.4 × 10 7 to 3.5 × 10 9 CFU cm −2 with the exception of one sample (V15) revealing a much lower abundance of mesophilic aerobic bacteria and a lack of Gram-negative bacteria. All other smear samples contained Gram-negative bacteria in a broad range from the detection limit of 6.3 × 10 1 CFU cm −2 up to 5.1 × 10 5 CFU cm −2 . With the exception of surface smear samples V9 and V13, all samples that contained Gram-negative bacteria harbored members of Enterobacterales in the range of 1.6 × 10 2 CFU cm −2 to 7.8 × 10 5 CFU cm −2 . Thus, in contrast to counts for total mesophilic aerobic bacteria, the abundance of Enterobacterales and Gram-negative bacteria varied strongly in the samples of different cheese varieties. smear samples contained Gram-negative bacteria in a broad range from the detection limit of 6.3 × 10 1 CFU cm −2 up to 5.1 × 10 5 CFU cm −2 . With the exception of surface smear samples V9 and V13, all samples that contained Gram-negative bacteria harbored members of Enterobacterales in the range of 1.6 × 10 2 CFU cm −2 to 7.8 × 10 5 CFU cm −2 . Thus, in contrast to counts for total mesophilic aerobic bacteria, the abundance of Enterobacterales and Gramnegative bacteria varied strongly in the samples of different cheese varieties.

Genotypic Identification of Gram-Negative Isolates by Partial Sequencing of 16S rRNA and gyrB Genes
A total of 85 Gram-negative isolates from the surface smear of the 15 surface-ripened cheese varieties were further processed for molecular identification based on partial 16S rDNA and gyrB sequencing (Table 2). Overall, the results revealed a high diversity comprising 22 different species that belonged to 14 different genera of Proteobacteria. Most isolates could be assigned to the γ-Proteobacteria class and 84.7% of all isolates represented Enterobacterales, whereas Xanthomonadaceae were represented by single isolates. Representatives of β-Proteobacteria not belonging to the order Enterobacterales comprised mainly Alcaligenaceae with 9.4% of total isolates and a single isolate of Uruburuella suis belonging to Neisseriaceae. Further single isolates represented members of Caulobacteraceae and Phyllobacteriaceae, which belong to α-Proteobacteria. Eight of the fifteen cheese varieties exclusively harbored members of Enterobacterales, whereas four cheese varieties additionally featured Alcaligenaceae or Neisseriaceae. The most abundant Gram-negative genera have been Proteus and Morganella with 37.7% and 23.5%, respectively, represented by the species Proteus vulgaris, Proteus hauseri and Morganella morganii. Proteus spp. were isolated from 11 of the 15 cheese varieties investigated in this study, and Morganella spp. were detected in 9 of the 15 cheese varieties ( Table 2). Further frequent isolates corresponded to Citrobacter sp.

Genotypic Identification of Gram-Negative Isolates by Partial Sequencing of 16S rRNA and gyrB Genes
A total of 85 Gram-negative isolates from the surface smear of the 15 surface-ripened cheese varieties were further processed for molecular identification based on partial 16S rDNA and gyrB sequencing (Table 2). Overall, the results revealed a high diversity comprising 22 different species that belonged to 14 different genera of Proteobacteria. Most isolates could be assigned to the γ-Proteobacteria class and 84.7% of all isolates represented Enterobacterales, whereas Xanthomonadaceae were represented by single isolates. Representatives of β-Proteobacteria not belonging to the order Enterobacterales comprised mainly Alcaligenaceae with 9.4% of total isolates and a single isolate of Uruburuella suis belonging to Neisseriaceae. Further single isolates represented members of Caulobacteraceae and Phyllobacteriaceae, which belong to α-Proteobacteria. Eight of the fifteen cheese varieties exclusively harbored members of Enterobacterales, whereas four cheese varieties additionally featured Alcaligenaceae or Neisseriaceae. The most abundant Gram-negative genera have been Proteus and Morganella with 37.7% and 23.5%, respectively, represented by the species Proteus vulgaris, Proteus hauseri and Morganella morganii. Proteus spp. were isolated from 11 of the 15 cheese varieties investigated in this study, and Morganella spp. were detected in 9 of the 15 cheese varieties ( Table 2). Further frequent isolates corresponded to Citrobacter sp.   Only two cheese varieties (V9 and V13) that harbored Gram-negative bacteria did not reveal a predomination of Enterobacterales. Cheese variety V9 harbored Stenotrophomonas maltophilia as the sole representative of Proteobacteria, whereas the surface microbiota of cheese variety V13 revealed a wide variety of different Proteobacteria, including Advenella sp., Advenella kashmirensis, Brevundimonas diminuta, Defluvibacter, Pusillimonas and Stenotrophomonas. In general, the numbers for different genera of Proteobacteria determined in the surface smear samples varied between 10 1 CFU cm −2 for the genera Brevundimonas and Defluvibacter and 10 4 CFU cm −2 for Advenella, Citrobacter, Morganella, Proteus, Providencia and Pusillimonas. Enterobacter, Hafnia, Kerstersia, Serratia, Stenotrophomonas and Uruburuella were represented with 10 2 to 10 3 CFU cm −2 .

Proteolytic and Lipolytic Activities of Isolated Proteobacteria
Many representatives of Enterobacterales are known to contribute to the formation of key aroma compounds in cheese by their enzymatic activity [34]. Consequently, 14 selected surface smear isolates from cheese varieties V2, V3 and V5 were tested for their proteolytic and lipolytic potential by applying a spot-on-the-lawn test. Only Proteus isolates revealed proteolytic activity in situ on Caso Bouillon Skim Milk Agar (Table 3) with isolates V2.4, V2.8 and V5. 3 showing highest proteolytic activity, as the whole CBSMA Agar plate was cleared at the latest after 5 days of incubation. Three Proteus isolates (V2.5, V2.8 and V5.3) revealed proteolytic activity also under anaerobic conditions, but only after 12 d of incubation (data not shown).
Lipolytic activity of isolates was tested on Tributyrin Agar (TBA). In general, lipolytic activity was much less frequently observed than proteolytic activity.
Halomonas and Proteus isolates from cheese sample V2 revealed lipolytic activity after 12 days' aerobic incubation at room temperature and 30 • C, whereas a Providencia heimbachae isolate from cheese sample V5 showed lipolytic activity in vitro after 12 days at 30 • C. No lipolytic activity could be observed under anaerobic growth conditions.

Volatile Carbonic Flavor Compounds Produced by Gram-Negative Isolates of the Cheese Surface Smear
To investigate the potential contribution of Proteobacteria to the typical flavor of semihard surface-ripened cheese, a headspace atmosphere analysis was performed on cheese variety V2, which was known to harbor different Proteobacteria such as Advenella, Morganella, Proteus, Providencia and Pusillimonas. Application of HS-SPME-GC-MS and HS-SPME-GC/PFPD revealed a variety of different volatile carbonic flavor compounds such as alcohols, aldehydes, carboxylic acids, ketones and sulfur compounds (data not shown). The majority of aroma compounds detected by the analysis were sulfur compounds (dimethyl sulfide, dimethyl disulfide, dimethyl trisulfide, hydrogen sulfide, methanethiol and methylthioacetate), followed by carboxylic acids (acetic acid, butyric acid, formic acid and 3-methylbutyric acid) and ketones (acetoin, 2-butanone, 2-nonanone and 1-phenylethanone). Alcohol compounds determined in the headspace comprised 3-methylbutanol and phenylethanol. Interestingly, many of the volatile organic compounds detected in significant amounts by headspace atmosphere analysis of cheese variety V2 are known to be produced by representatives of Enterobacterales [14,16,19,[35][36][37][38][39][40].

Extended Spectrum β-Lactamase (ESBL)-Mediated Resistances of Enterobacterales Isolated from the Cheese Surface Smear
To investigate potential health risk factors of individual smear bacteria, selected isolates that come into consideration as ESBL producers were tested for their antibiotic susceptibility and the ability to produce extended spectrum β-lactamase. However, none of the Morganella, Enterobacter, Citrobacter, Proteus, Hafnia and Serratia isolates featured ESBL production, whereas all the Morganella isolates as well as one isolate each of the genera Enterobacter, Citrobacter and Hafnia expressed AmpC β-lactamase, which is intrinsically present in these species.

Discussion
The present study revealed an unexpectedly high abundance and diversity of Proteobacteria species on the surface of different semi-hard surface-ripened cheese varieties at the end of the ripening stage. The composition of Gram-negative bacterial population of the surface microbiota varied considerably between different cheese varieties. High counts of up to 7.8 × 10 5 CFU cm −2 Gram-negative bacteria were determined for the majority of cheese varieties analyzed in the present study. These values were comparable to counts of approx. 10 5 CFU cm −2 Gram-negative bacteria determined for Tilsit-type cheeses as reported in the literature for [5,13]. Differences in the quantity of Gram-negative bacteria might be influenced by differing production technologies, seasonal variation or the milk type applied for cheese production. With up to 10 9 CFU cm −2 , the quantity of total aerobic mesophilic bacteria in cheese smear was very similar for all cheese varieties analyzed in this study, and corresponded well to values reported for various surface-ripened cheese varieties in the literature [3,5,6,8].
At least 22 different species of Proteobacteria were isolated, belonging to 14 different genera. With 84.7% of the 85 identified isolates, Enterobacterales represented the most frequent Gram-negative bacteria.
About half of the analyzed cheese varieties featured exclusively members of Enterobacterales, whereas only two varieties lacked its representatives. The observed predomination of Enterobacterales as major representatives of Gram-negative bacteria isolated from cheese surface microbiota is consistent with data from previous studies on surface-ripened cheese [12,41]. In this study, Enterobacterales were mainly represented by isolates of the genera Proteus and Morganella, whereas Citrobacter, Enterobacter, Serratia, Providencia and Hafnia were isolated to a much lesser extent. Besides Enterobacterales, Gram-negative bacteria regularly detected on the surface of surface-ripened cheeses belonged mainly to Proteobacteria families such as Alcaligenaceae, Caulobacteraceae, Halomonadaceae, Moraxellaceae, Oceanospirillaceae, Pseudoalteromonadaceae, Pseudomonadaceae, Vibrionaceae and Xanthomonadaceae [12,13,15,20,24,[42][43][44]. On Livarot, a French surface-ripened soft cheese, Gram-negative bacteria accounted for 32% of the overall bacteria isolated from the cheese surface [15]. A recent study on the temporal differences in the microbial composition of Époisses cheese rinds during ripening and storage also revealed a dominance of Gram-negative species. At the end of ripening, about 70% of all 16S rRNA gene sequences were derived from Gram-negative bacteria, with the most abundant genera being Psychrobacter, Halomonas, Mesonia and Vibrio [28].
Proteobacteria other than Enterobacterales isolated from the surface of cheese varieties investigated in this study were mainly represented by single isolates, and comprised members of Alcaligenaceae (Advenella, Kerstersia and Pusillimonas) as well as Caulobacteraceae (Brevundimonas), Phyllobacteriaceae (Defluvibacter), Neisseriaceae (Uruburuella) and Xanthomonadaceae (Stenotrophomonas). Brevundimonas diminuta and Stenotrophomonas sp. were already reported to be part of the surface smear microbiota of different surface-ripened French cheeses [12]. Advenella kashmirensis as well as Pusillimonas sp. have been described for Austrian mountain cheese [45]. To our best knowledge, for the genera Defluvibacter, Kerstersia and Uruburuella this is the first description of an association with cheese. Defluvibacter sp. represents typical environmental bacteria isolated from aqueous habitats such as the activated sludge of waste water plants [46]. In contrast, Kerstersia gyiorum has been isolated from various human clinical samples such as wounds, feces or chronic infection sites [47]. Although the definite pathogenic potential has not been reported for this species, its contribution to chronic infections is discussed [47]. Likewise, Uruburuella suis, isolated first from clinical specimens of animals, seems to be related to respiratory diseases in pigs [48].
Our study revealed no correlation between the abundance of Gram-negative bacteria or species diversity and the type of milk applied for cheese production, whether raw, thermized or pasteurized. This is consistent with the finding that on an average, similar percentages of Proteobacteria were determined in raw milk and pasteurized milk cheese using high-throughput sequencing analysis of 62 different cheese samples [20]. Furthermore, the type of production, whether artisanal manufacturing, traditional or mass production, did not influence the quantity and diversity of Proteobacteria. It is reported that members of Proteobacteria populate functional habitats such as milk handling surfaces or cheese maturation surfaces in artisanal cheese manufacturing plants [24]. In particular, coliform bacteria and Pseudomonadaceae are known to be part of biofilms on wooden vats and shelves from which inoculation might take place during cheese production or ripening [25]. Halotolerant Proteobacteria such as Psychrobacter, Pseudoalteromonas and Vibrio are supposed to establish a 'house microbiota', as they are similarly abundant in aging rooms and cheese surfaces [24]. While cheeses produced from pasteurized milk might exclusively be colonized by a secondary contamination with Proteobacteria, those produced from raw or thermized milk might contain indigenous Proteobacteria originating from the milk. Raw milk can be populated by significant proportions of mainly psychrotrophic Proteobacteria (Enterobacteriaceae, Pseudomonas, Acinetobacter) that might flourish during cold storage, and even dominate dairy tank milk [27,42]. As Gram-negative bacteria are known to be adherent to stainless steel and therefore exhibit a high biofilm-forming potential, they are able to colonize dairy pipelines leading to potential recontamination of pasteurized milk [26].
However, for San Simón cheese it was demonstrated that counts for Enterobacteriaceae in the surface smear constantly decreased during a 6-week ripening period, which was related to the influence of the combined effects of physico-chemical parameters [19].
It is assumed that members of Enterobacterales contribute to typical cheese flavor development by the lipolytic breakdown of milk fat and the generation of semi-volatile fatty acids and the according ethyl esters, while peptidase and deaminase activity on proteins contributes to the production of volatile aroma compounds [35,40]. In this study, lipolytic activity was observed for isolates of Halomonas sp., Proteus sp., and Providencia heimbachae, which seems to be the first description of lipolytic activity for Halomonas and Providencia. According to the literature, Citrobacter spp. and Enterobacter spp. may also contribute to aroma compound production in cheese although their proteolytic and/or lipolytic activity in vitro seems to be less pronounced [49].
Proteus spp., in particular Proteus vulgaris strains, were frequently isolated from the surface of surface-ripened cheeses, and their potential role in cheese ripening has already been intensively discussed [12][13][14][15]36,40,50]. Proteus vulgaris was able to successfully colonize and dominate the surface of pilot-scale cheese while significantly contributing to its organoleptic properties through the production of aldehydes [14]. Therefore, the presence of Proteus vulgaris might be desirable for the development of typical aroma compounds [50]. In agreement with the high caseinolytic activity reported for crude culture supernatants of Proteus vulgaris, all Proteus strains investigated in this study featured strong proteolytic activity [40]. In this study, Enterobacterales isolates other than Proteus spp. showed no casein degradation, although this was described for cheese smear isolates affiliated to the genera Enterobacter or Citrobacter [16,34,49]. Accordingly, this enzymatic activity seems to be a strain-dependent feature [34]. In experimental cheese produced from milk artificially inoculated with Enterobacteriaceae, the degradation of major casein fractions could be assigned to their specific proteolytic activity [34].
Enzyme-driven catabolic activities such as proteolysis and the subsequent catabolism of amino acids or lipolysis, besides lactose and citrate metabolism implying glycolysis, impact to a great extent the organoleptic properties of cheese [51,52]. In the literature, various Proteobacteria species are described to produce a wide variety of volatile compounds in large quantities when growing in milk, cheese model medium or cheese surface smear, thereby impacting cheese flavor during ripening [14,16,36,37,40]. A variety of the aroma compounds typical for cheese flavor were detected in significant amounts in the volatile compound pattern of the headspace atmosphere of cheese variety V2 analyzed in this study. Mainly volatile compounds such as carboxylic acids, ketones and sulfur compounds could be related to Enterobacterales isolates after comparing the results to bibliographic data. As Morganella and Proteus spp. predominated the Gram-negative fraction of isolates from the surface smear of cheese variety V2, we concluded that they may contribute to the respective volatile aroma compounds detected in the headspace atmosphere of cheese variety V2. In particular, Proteus vulgaris is known to produce a wide variety of volatile compounds with low perception threshold in large amounts (dimethyl disulfide, 3-methylbutanol) in cheese model medium, and has a prevalence on aroma compound production during the ripening of a model cheese [14,40]. A Proteus vulgaris isolate from the surface of surfaceripened French cheese was able to produce inter alia dimethyl-trisulfide, dimethyl-sulfide, 2-butanone, 1-phenylethanone and phenylethanol in liquid medium [36]. These volatile compounds were also detected in the headspace of cheese variety V2, providing further evidence that Proteus sp. may contribute to the formation of the volatile sulfur compounds that represent garlic odor notes known to be increasingly produced at the end of the ripening process [40]. This assumption is supported by a study which indicates that using human senses (odor and taste), it is possible to predict high levels of Enterobacteriaceae in soft cheeses made from raw milk [53].
As Proteus vulgaris and Hafnia alvei produce a wide variety and large quantity of volatile compounds, non-pathogenic and non-biogenic amine-producing strains thereof were proposed as industrially applicable microorganisms for the production of natural cheese flavor compounds [49]. Hafnia alvei inoculated in pilot-scale cheese was able to establish successfully at the beginning of the ripening process in the cheese community and clearly intensified the production of volatile compounds, mainly by production of volatile sulfur compounds [37]. Hence, non-pathogenic strains of Hafnia alvei are commercially available for the purpose of conferring a flavor comparable to raw milk cheese to that produced from pasteurized milk [37]. Accordingly, as one of few Enterobacterales, Hafnia alvei was included in an authoritative list of microbial food cultures with practical use by the International Dairy Federation (IDF) [54]. However, the potential for production of biogenic amines in vitro or the presence of antibiotic resistance genes might question the use of such bacteria in cheese production [12,16,17].
Concerning the presence of Proteobacteria in food products such as cheese, the positive as well as negative aspects considering the risks and benefits of Gram-negative bacteria in cheese surface smear are up for debate. Many traditional cheeses harbor Proteobacteria, which are considered to belong to natural autochthonous microbiota and are thought to have a beneficial impact on cheese sensory characteristics [14][15][16][17]41]. However, a previous study on the development of a defective smear after prepackaging of surface-ripened cheese in plastic foil revealed that the presence of certain members of Proteobacteria may take a turn for the worse. The vacuum foil prepackaging of surface-ripened cheese can result in a shift in the microbial composition of the smear microbiota towards Proteobacteria, which would then contribute to the development of negative organoleptic properties in the surface smear of the cheese [55]. Furthermore, certain members of Enterobacterales found in cheese are regarded as markers for unsatisfactory hygiene practice during food processing due to their relation to bacterial strains of fecal origin [56]. Threshold and tolerance values for the number of indicator microorganisms revealing bacteriological safety exist in the food legislation of almost all European countries.
In the case of Swiss cheese, Swiss food hygiene legislation prescribes limit values for Escherichia coli in cheese produced from heat-treated milk but not for raw milk cheeses. However, limit values are defined for Salmonella spp. as food safety criteria in cheeses produced from raw milk or milk with a heat treatment below pasteurization temperature (thermized milk). Accordingly, as no Escherichia coli or Salmonella spp. were isolated from the cheese surface, the occurrence of Citrobacter, Enterobacter, Hafnia, Morganella, Proteus, Providencia and Serratia in cheese smear is in agreement with current regulations.
Beside hygiene-related aspects, food safety issues challenge the occurrence of Gramnegative bacteria in cheese. According to the European Food Safety Authority (EFSA), none of the Gram-negative bacterial species identified in this study qualified for presumption of safety (QPS) status when intended as biological culture added to foods [57]. Food-borne outbreaks related to Gram-negative bacteria in cheese are rare, and cover mainly Salmonella, enteropathogenic or shiga toxin-producing E. coli or Brucella spp. [58]. Intense hygienic efforts and improved technology applied in the dairy process, such as pasteurization and proper acidification of milk during cheese production as well as the preservation of semi-hard cheese by salting, minimize the risk for transfer of food-borne pathogens. Furthermore, certain strains of Gram-negative species exhibited an antilisterial effect in vitro and in situ in previous studies [59]. Though several Proteobacteria species isolated in this study (Brevundimonas diminuta, Citrobacter sp., Enterobacter sp., Hafnia alvei, Morganella morganii, Proteus vulgaris, Stenotrophomonas maltophilia, Serratia proteamaculans) represent opportunistic pathogens (classified to biosafety level 2), they are associated with diseases that are rarely serious.
Antibiotic resistances such as extended spectrum β-lactamase (ESBL)-mediated resistance of Enterobacteriaceae concern a further safety issue. Several isolates of Enterobacterales from this study have been tested for extended spectrum β-lactamase (ESBL) production, but none of the isolates exhibited such a resistance. The detected AmpC β-lactamases are intrinsically present in many species. They are chromosomally encoded and usually expressed at low levels [60,61]. Considering the literature about ESBL producing Enterobacteriaceae related to cheese, potential β-lactam resistance was detected mainly in Escherichia coli strains isolated from cheese produced in different countries [60]. As the prevalence of ESBL in Europe is high, the emergence of antibiotic resistance in food-related bacteria needs to be closely monitored [62]. Although the consumption of antibiotic-resistant Enterobacteriaceae may not present an immediate risk to human health, antibiotic-resistant genes might be transferred from commensal bacteria to obligate pathogenic bacteria on the cheese surface or bacteria of the human intestinal microbiota [12].

Conclusions
This study provides valuable insights into the occurrence of Enterobacterales in cheese surface microbiota and their presumable contribution to flavor development in cheese. An unexpectedly high diversity of Gram-negative bacteria was detected in the smears of different semi-hard surface-ripened cheese varieties. Proteus spp. and Morganella spp. were most prominent, independent of the milk type applied for cheese production or the type of manufacturing process. The quantity of Gram-negative bacteria in the surface smear differed between different cheese varieties, and could have been influenced by dairy 'house microbiota'. Proteus spp. seemed to contribute in particular to the organoleptic properties of surface-ripened cheese by their strong proteolytic activity. Furthermore, a variety of volatile flavor compounds with low perception threshold was detected in the headspace atmosphere of the surface of cheese variety V2, which most likely resulted from the enzymatic activity of Enterobacterales. On the other hand, several isolated Enterobacterales species represented opportunistic pathogens. Therefore, safety issues related to these species need to be investigated in order to manage the risk to consumer health.
In conclusion, cheese manufacturers are encouraged to produce as tasteful a cheese as possible, with an appealing appearance and of high quality, stability and safety, independent of the seasonal variability induced in basic raw material. Autochthonous-occurring Enterobacterales on the cheese surface appear to be valuable for the organoleptic properties of various cheese varieties, and must not entirely be considered as undesired contaminants. Naturally occurring surface smear microbiota comprising Enterobacterales may protect the surface against unwanted mold spoilage or pathogenic contaminants. Consequently, abandonment of Enterobacterales from the surface smear due to food safety reasons might result in cheese lacking the typical flavor that is important for the authenticity of its variety.