Twenty cheese samples, corresponding to six different types of Portuguese soft and semi-soft ripened cheeses were used. Table 1 shows the origin and the occurrence of thermal processing of the milk. The cheeses were random collected in 2007, at retail establishments in different regions, from the North to the South of Portugal. From each cheese, an aliquot of 25 g (which included approximately equal amounts of material from the inner and outer parts of the cheese) was removed and blended for 2 min, in a Stomacher^® (Barcelona, Spain), with 225 mL of buffered peptone water (BPW) (pH 7.0) supplemented with 50 μg/mL cycloheximidine (Sigma, Saint Louis, USA).

The enumeration of AMP^r bacteria was performed on five cheese samples randomly selected from five types of cheese (A, B, C, D and F). Portions of 0.1 mL of the homogenised or of the appropriately diluted cheese samples in BPW containing 50 μg/mL cycloheximidine, obtained as described above, were immediately surface plated onto three replicate plates of VRBG (Violet Red Bile Glucose) agar (Oxoid, Hampshire, England) supplemented with 20 μg/mL of sodium ampicillin (AppliChem, Damstadt, Germany). Plates were incubated aerobically at 37 ºC for 38 h before counting the colonies. Although this medium is used to select for Enterobacteriaceae, all the AMP^r isolates were counted.

In order to recover all the resistance phenotypes present in the cheeses, even stressed ones due to milk and/or cheese processing, after sampling each one of the twenty homogenised cheese samples were overnight incubated at 37 ºC, before surface plating on VRBG and on SS (Salmonella & Shigella) agar (Oxoid, Hampshire, England). Both media were supplemented with 20 μg/mL of sodium ampicillin (AppliChem, Damstadt, Germany). Plates were incubated aerobically at 37 ºC for 24 to 48 h. Different morphological types of AMP^r colonies were selected and replicated five times for phenotypic confirmation. Strains Escherichia coli XL1-Blue and Escherichia coli XL1-Blue transformed with plasmid pUC18 (which contains the TEM-1 β-lactamase gene) were used as negative and positive controls, respectively for AMP^r phenotype identification.

The detection of β-lactamases produced by the AMP^r isolates was performed by adding nitrocefin (Oxoid, Hants, England) to the cell biomass, according to the manufacturer instructions. For the positive nitrocefin strains, the susceptibility pattern determination was performed by the disk diffusion method (DDM) on Mueller Hinton agar (Oxoid, Hampshire, England) with antibiotic disks (Oxoid, Hampshire, England) according to the Clinical Laboratory Standards Institute (CLSI), formerly National Centre Clinical Laboratory Standards (NCCLS) [15,16]. Thirteen different antibiotics were used: amoxicillin/clavulanic acid combination (Clavamox) (AMC) 30 μg/10 μg, respectively; ceftazidime (CAZ) 30 μg; cefotaxime (CTX) 30 μg; cefpirome (CPO) 30 μg; aztreonam (ATM) 30 μg; cefoxitin (FOX) 30 μg; imipenem (IPM) 10 μg; meropenem (MEM) 10 μg; chloramphenicol (CHL) 30 μg; tetracycline (TET) 30 μg; gentamicin (GEN) 10 μg; trimethoprim/sulfamethoxazol (Bactrim®; SXT) combination (1:19) and ciprofloxacin (CIP) 5 μg.

In order to observe the influence of clavulanic acid on both CAZ and CTX, these antibiotics were positioned in line, with (AMC) in the middle. The results of the antimicrobial susceptibilities were interpreted according to the CLSI guidelines.

The identification of sixty β-lactamase producing isolates randomly selected from VRBG and SS plates was carried out in two steps. Firstly, classical morphological and physiological tests were performed in agreement with the Bergey’s Manual of Determinative Bacteriology [17]. These tests included Gram staining reaction, morphology, motility, and biochemical tests including: fermentative/oxidative metabolism of glucose [18] (fermentation of glucose to acid or to acid and gas), aerobic/anaerobic growth, catalase and oxidase reactions. Subsequently, the isolates identified as belonging to the Enterobacteriaceae family were submitted to additional identification by using the miniaturized Enterosystem 18R galleries (Liofilchem Bacteriology Products, Roseto, Italy), following the manufacturer instructions. Enterosystem 18 R system is based on the metabolism of orthonitrophenol (ONPG), lysine, ornithine, arginine-decarboxylase, phenilalanine-deaminase, citrate, malonate, urease test, H₂S, and Voges-Proskauer reaction, indol test and in the fermentation of glucose, mannitol, inositol, sorbitol, sucrose, arabinose and raffinose.

The identification of the genes involved in β-lactam resistance was performed, in 21 random selected β-lactamase producing isolates, by a multiplex-PCR approach for three β-lactamase (bla) genes (bla_TEM, bla_SHV and bla_CTX-M) according to Monstein et al. [19], except that genomic DNA was extracted by the guanidine thiocyanate method according to Pitcher et al. [20].

Strains E. coli UB0402407 and K. pneumoniae HY0301692, kindly provided by the Clinical Microbiology Department, Linköping University, Sweden, were used as positive controls for bla_TEM + bla_CTX-M genes and bla_SHV gene, respectively.

The enumeration of ampicillin resistant bacteria was performed on VRBG agar, a medium for the detection and enumeration of bile-tolerant Enterobacteriaceae in dairy products and foods, that is becoming a preferred medium for use in investigations into raw materials, processed foods and plant hygiene. Figure 1 shows the counting data of AMP^r bacteria on VRBG plates, relative to five cheeses samples corresponding to five different types of cheese (A, B, C, D and F, Table 1). According to the type of cheese, the numbers of AMP^r varied from about 4.7 × 10² CFU/g to 1.5 × 10⁷ CFU/g.

Although Enterobacteriacea are used as a parameter in process hygiene criteria for various food products, current legislation in Europe does not use Enterobacteriaceae as microbiological criteria for cheese made from raw milk. Nevertheless, in cheeses made from pasteurized milk the level of contamination with E. coli may not exceed 10³ CFU/g [21]. Various genera of Enterobacteriaceae are commonly found in milk and cheese [4], although the highest counts close to 10⁷ CFU/g were generally found during the first days of ripening, decreasing afterwards [22]. In this work, the numbers of AMP^r bacteria varied according to the type of cheese (Figure 1). The lowest values were obtained for cheese types A (3.0 × 10³ CFU/g) and D (4.7 × 10² CFU/g). The first is a DOP (Denomination of Protected Origin) cheese made from raw ewe’s milk, the second is a cheese made from pasteurized goat’s milk (Table 1). The results from the counts in cheese type A are an example that it is possible to produce good quality cheese with raw milk. The results from cheese type D, made from pasteurized goat’s milk confirm that the thermal processing of the milk is an important factor to decrease bacterial contaminations. Nevertheless, to rely only on pasteurization to solve bacterial contaminations is likely to limit efforts to improve total hygiene in cheese production.

The highest level was obtained for type C cheese made from raw goat’s milk (1.5 × 10⁷ CFU/g). Since the majority (85%) of the isolates identified belonged to the Enterobacteriaceae family, the suggested level of contamination of the cheese from type C, deduced from the presented data, is concerning.

In order to avoid high rates of cheese contamination with these and other potential hazardous bacteria, the existence of good sanitary conditions during milking and cheesemaking is obligatory. This includes herd health management, particularly udder health preventive management, hygiene of milking, promptly and properly refrigeration of the milk after milking and hygiene along the cheese production chain.

The detection of AMP^r bacteria in cheeses was performed by using VRBG agar and SS agar. In both media, the growth of gram-positive microorganisms is inhibited due to the presence of either pure bile salts, mixtures of bile salts or dyes. SS agar has been developed for the selection and differentiation of enteric microorganisms, from clinical and nonclinical materials.

A total of 182 AMP^r selected isolates recovered from VRBG and from SS plates were screened for the presence of β-lactamases by the nitrocefin test. From the 182 isolates, 172 were found positive (102 from VRBG and 70 from SS plates) indicating the presence of β-lactamases in 94.5% of the AMP^r isolates. The three isolates selected from cheese type E (made from pasteurized cow’s milk) were within the 10 nitrocefin negative isolates.

Concerning the suitability of both media to recover β-lactam-resistant bacteria, the results showed that both media fulfilled this objective. From a total of sixty β-lactamase producers submitted to identification, only three (6.7% of the isolates) and five (15% of the isolates) species were recovered only from SS or from VRBG plates, respectively (Table 2).

Sixty nitrocefin positive isolates (37 from VRBG and 23 from SS plates) were identified as belonging to the Enterobacteriaceae family (85%) and to the genera Pseudomonas (8.3%) and Aeromonas (6.7%) (Table 2). Within the Enterobacteriaceae family, 34 isolates (12 from SS and 22 from VRBG plates) were identified as belonging to eleven different species. It was not possible to identify at species level 17 Enterobacteriaceae isolates (28.3%) recovered from both media. The type cheese B, made from raw ewe’s milk (Table 1) was the source of most of the identified isolates. However 13 out of the 20 cheeses analyzed were from this type. Five different enterobacterial species were collected from type cheese F (made from a raw milk mixture from cow, ewe and goat) although a single cheese was analyzed. Regarding the recovery medium, from the eleven identified species, five (Enterobacter cloacae, Proteus mirabilis, Citrobacter freundii, Enterobacter sakasakii and Enterobacter alvei) and three (Serratia rubidaea, Providencia retgeri and Shigella boydii) species were only collected from VRBG or SS plates, respectively (Table 2). Pseudomonas and Aeromonas isolates were recovered from both media.

The results from antimicrobial susceptibility determination for the 13 antimicrobial agents and the respective percentage of the resistance phenotypes obtained showed a higher resistance to AMC (51.7%) and FOX (46.5%) and to TET (46.5%) (Table 3).

It is also noticed a high resistance level (38.4 %) to SXT. Nevertheless, antimicrobial agents which seem more effective, thus having higher susceptibility among these bacterial strains are GEN (98.3%), CIP (97.7%), IPM (95.3%) and MEM (90.7%) (Table 3). In this study, a multidrug resistance (MDR) phenotype was defined for isolates that were resistant to two or more structurally unrelated antimicrobial agents. According to this, 31.4% of the different phenotypes detected displayed a MDR phenotype (Table 4). The most prevalent resistant phenotypes have patterns of SXT and/or TET plus AMC and FOX and finally CHL.

The most effective antimicrobial agents against the bacterial strains isolated are GEN (98.3%), CIP (97.7%), IPM (95.3%) and MEM (90.7%). It seems noteworthy the high bacterial susceptibility to CIP found since in Portugal, in contrast to other European countries, this type of antibiotics is one of the most frequently used [23]. The high frequency of resistance to AMC (51.7%) might be due to the their extensive usage in Europe, since penicillins represent the most widely used antibiotic class in all 25 participating countries studied by Ferech et al. [24] in the ESAC Project Group. These authors reported a distinct shift from narrow-spectrum penicillins to broad-spectrum penicillins, and specifically their combinations with β-lactamase inhibitors, during the period 1997–2003.

On the other hand, the equally high resistance to FOX (46.5%) a cephamycin, is probably because it has been on the market for a longer period which may have allowed for the development of more resistance than those that are more recently available [25]. The shifting in antibiotic use in Europe is extremely high for this class of antibiotics, suggesting that in many European countries these antibiotics are prescribed inappropriately [25].

The high resistance detected to TET (46.5%) may be related with their extensive use in veterinary, [26–29] including aquacultures. Nevertheless, it could also mean an eventual non efficiency in humans through selection, following the widespread use of these antibiotics [27].

The most prevalent resistance phenotype is AMP + TET (11.0%) followed by the MDR phenotypes AMP + SXT + TET and AMP + AMC + FOX + SXT + TET (each 4.7%). This fact might result from the association of TET with the massive use of Bactrim® (SXT) [30] mainly for the treatment of urinary infections [31]. Other sulfonamides are also described as widely used for both veterinary and aquacultures purposes [27]. Among the MDR phenotypes, resistance to CHL was also present in important phenotypes such as AMP + AMC + FOX + CHL + TET (4.1%) and AMP + AMC + FOX +CHL + TET + SXT (2.3%) or AMP + AMC + FOX + CHL + SXT (2.3%).

The relatively high number (31.4%) of MDR phenotypes found among bacteria isolated from different types of cheese, may contribute to the emergence of a serious public health problem. These finding are in agreement with the results of Fernandes et al. [32], that reported the spreading of MDR in hospitals and in the community by isolates of E. coli coding for ESBLs. The same effect was also reported in other European countries [32,33].

Twenty one randomly selected isolates were tested for the presence of three β-lactamases genes. The expected sizes of these amplicons, according to the control strains used, were 445 bp for the bla_TEM, 593 bp for the bla_CTX-M and 747 bp for the bla_SHV. The multiplex-PCR analysis showed that, while the bla_CTX-M and bla_SHV were absent, 80.9% of the tested isolates harbor a gene for the TEM-type enzyme (bla_TEM).

The presence of bla_TEM in bacteria isolated from cheese reported in this study, is in agreement with several recent reports showing that TEM β-lactamases are particularly involved in resistance phenotypes presented by bacteria isolated from several agro-industrial environments such as in water reservoirs [34], wastewaters [35], livestock [36–39] and even in transgenic plant fields [40]. The presence of TEM β-lactamases in these bacteria may be explained by its insertion between transposable elements that move rapidly through bacterial populations [41] and may become a major public health concern. However, TEM-β-lactamases alone may not explain all the resistance phenotypes found in the present study. Regarding the results presented in Table 3, they may be explained by several other mechanisms that include the presence of other β-lactamases, alteration of membrane permeability such as due the presence of a porin loss, modification of target, efflux mechanisms. In what concerns to the mechanisms related to other antimicrobial rather than β-lactams the mechanisms that may explain these resistance phenotype are identical [5,6]. In order to clarify this subject, further studies are needed.