Pathogens in the Food Chain: Escherichia coli Strains in Raw Milk Originating from Ewes Treated for Mastitis with Various Therapeutic Protocols

: Dairy products from ovine milk are very popular in the Mediterranean countries and are gaining a large portion of the market in EU countries and worldwide. EU legislation permits the dairy processing of raw ovine milk under certain conditions. To study the ecology and prevalence of E. coli in raw ewes’ milk and assess thus the public health risk, samples of milk were taken from 75 different sheep farms in the rural area of Epirus, Greece. The initial sampling was conducted in clinically healthy animals which were noted as controls (group A). From the same farms, samples were taken from animals with clinical mastitis and before treatment (group B). For therapeutic purposes, to some animals, a combination of penicillin and streptomycin was administrated (group C1), or tetracycline (group C2), or enrofloxacin (group C3). Finally, samples of raw milk were taken from the C groups, on the first day after the withdrawal period of the antibiotics used, when the milk is permitted to enter the food chain. In total, 97 isolates of Escherichia coli were recovered from all groups. Analysis revealed an impressive increase in E. coli strains in the milk of group B (39.33%) with respect to group A (5%). Even after treatment, although the prevalence was decreased, it was still found to be higher in the C groups than group A. E. coli O157:H7 strains absent from group A were detected in all other groups in relatively low occurrence rates with respect to other “O” serotypes but non-O157:H7 strains. Virulence factors such as the production of toxins (32.60% for serotoxin 1 and 18.47% for serotoxin 2) and hemolysin (42.39%) as well as biofilm formation capacity (52.17% of the total) and ESDL production (43.47% of the total) were also studied. All strains were also tested for susceptibility against 12 antibiotics by the MIC method and the results showed a high prevalence of resistance and multi-resistance. The presence of various resistant strains to antibiotics and pathogenic “O” serotype strains in the milk when it can enter the food chain again is an alarming conclusion.


Introduction
In Greece, as in most Balkan and Mediterranean countries, ewes' milk has a significant financial importance.According to the EU, 57.7% of the milk delivered to the dairies in Greece originates from ewes [1].Ovine milk and particularly ovine dairy products have been important staple foods since antiquity.Ovine species are perfectly adapted to the Mediterranean climate and besides their milk, they also provide meat, skin, and wool, contributing thus significantly to the survival of the local populations in historical as well as contemporary times.As far as the safety of these dairy products is concerned, their consumption has been associated with various diseases [2][3][4], and this is not a surprise, given that milk is an excellent medium for the growth of almost all pathogens and that the hygienic status of many farms is often not acceptable.At the same time, there is a growing demand and preference for raw ovine milk and dairy products made from such milk [5].This movement combines these products which are usually produced by local small-scale dairies, with attractive concepts such as "natural", "green", "organic", "authentic taste", and "traditional" [6,7].In these establishments, however, the hygienic status is often compromised, thus increasing the odds of products detrimental to public health.
Milk, per se, is a complex matrix of ingredients and an excellent medium for bacteria, pathogenic or not.It has been argued that milk has its own defense systems against bacteria, such as its microbiota whose populations act antagonistically against pathogenic intruders [8,9].But even these defense systems succumb to various factors such as the environmental temperature, the environmental spoilage, the welfare of the animals, the sanitation of the milking apparatus and of the milking process, and the health of the udders and of the animals, to name a few [10][11][12][13][14][15][16][17][18].Contaminated milk in the bulk tank is a common finding [19][20][21].
Milking is the initial and most crucial stage of the dairy food chain since bacteria such as Escherichia coli can contaminate the milk and compromise the quality and safety of the final products [22,23].E. coli O157:H7 is a serotype of special significance due to its high virulence but other serotypes are also of high concern, such as the-so-called-"big six" (O26, O45, O103, O111, O121, and O145), owing their nickname to their association with severe foodborne illness [24].
Although EU legislation (EC 852/2004, EC 853/2004) enforces pasteurization [25,26] in raw milk, it permits under certain conditions the processing of raw milk without any thermal treatment, based on the total bacterial count (TBC), the absence of Brucella species from the milking animals and the ripening of the curd for a period greater than two months.Yet none of these conditions ensures the absence of pathogenic E. coli from the final dairy products.Many pathogens can survive for weeks or even months in dairy products.Salmonella spp, Staphylococcus aureus, Listeria mocytogenes, Coxiella brunetti, and Escherichia coli have been isolated from raw milk as well as from cheeses made from raw milk [27][28][29][30][31][32][33][34].
Clinical mastitis is a serious problem in milk-producing flocks and the major cause of "milk drop syndrome" [35].E. coli is involved in the pathogenesis of clinical mastitis along with other pathogens [36].In the early stages of mastitis, it is possible that milk from unnoticed sick animals may end up in the same bulk tank as the milk from the healthy ones.After successful treatment with antibiotics when the clinical signs have vanished, it is possible that some pathogens may remain in the udder contaminating the milk.From the perspective of food safety, it is of utmost importance to investigate the diversity and prevalence of the hazardous E. coli "O" serotypes in milk from healthy udders, in milk from udders suffering from mastitis, and in milk after the withdrawal period of the antibiotics used to treat mastitis, when milk enters the food chain again, and these are the aims of this study (Figure 1).

Description of Sampled Animal Groups
The study was conducted in 75 sheep farms located in the region of Epirus, G All farms were situated in a lowland area, sharing pastures with similar vegetation size of the flocks was 100-300 animals of a local breed ("boutsikes"), and all used the pastures and followed similar farming practices such as vaccinations, antiparasitic ments, use of milking machines, administration of concentrated feeds, etc.Samples o were collected from healthy sheep with healthy udders (referred exclusively to as th trol group-"martyrs"), from sheep with clinical mastitis before treatment (clinical ev tion points: color and temperature of udders, pain, change in color and viscosity of decreased milk production and California Mastitis Test category of grade 2 and hig and from the previous animals after treated for mastitis.All animals were in the m of their milking period and were randomly selected using tables of random num based on their ear tag numbers.The animals suffering from mastitis were kept away the healthy ones under strict hygienic conditions.Diagnosis and treatments were ca out by professional veterinarians.It is important to clarify that identifying the respon microorganism in clinical practice is usually not feasible, cost-effective, or practical d the remoteness of the farms and the mountainous terrain.Most veterinarians in G determine the treatment regimen empirically, based on clinical signs and the therap history of the flock (efficacy of previous treatments, success rates, etc.).Three grou animals were formed (Figure 1): (a) control animals with clinically healthy udders-g A-control; (b) animals suffering from mastitis, before treatment (infected udder)-g B; and (c) the animals of group B, at the first day after the end of the withdrawal per each antibiotic used (treated udder)-group C. The treated ewes were divided into subgroups, receiving different drastic compounds.The therapeutic protocols use cluded the following antibiotic substances: (1) penicillin and streptomycin, (2) oxyte cline, and (3) enrofloxacin (Table 1).On the first day after the end of the withdrawal p when the milk can be given to the dairies for processing, samples were taken fro

Description of Sampled Animal Groups
The study was conducted in 75 sheep farms located in the region of Epirus, Greece.All farms were situated in a lowland area, sharing pastures with similar vegetation.The size of the flocks was 100-300 animals of a local breed ("boutsikes"), and all used the same pastures and followed similar farming practices such as vaccinations, antiparasitic treatments, use of milking machines, administration of concentrated feeds, etc.Samples of milk were collected from healthy sheep with healthy udders (referred exclusively to as the control group-"martyrs"), from sheep with clinical mastitis before treatment (clinical evaluation points: color and temperature of udders, pain, change in color and viscosity of milk, decreased milk production and California Mastitis Test category of grade 2 and higher), and from the previous animals after treated for mastitis.All animals were in the middle of their milking period and were randomly selected using tables of random numbers based on their ear tag numbers.The animals suffering from mastitis were kept away from the healthy ones under strict hygienic conditions.Diagnosis and treatments were carried out by professional veterinarians.It is important to clarify that identifying the responsible microorganism in clinical practice is usually not feasible, cost-effective, or practical due to the remoteness of the farms and the mountainous terrain.Most veterinarians in Greece determine the treatment regimen empirically, based on clinical signs and the therapeutic history of the flock (efficacy of previous treatments, success rates, etc.).Three groups of animals were formed (Figure 1): (a) control animals with clinically healthy udders-group A-control; (b) animals suffering from mastitis, before treatment (infected udder)-group B; and (c) the animals of group B, at the first day after the end of the withdrawal period of each antibiotic used (treated udder)-group C. The treated ewes were divided into three subgroups, receiving different drastic compounds.The therapeutic protocols used included the following antibiotic substances: (1) penicillin and streptomycin, (2) oxytetracycline, and (3) enrofloxacin (Table 1).On the first day after the end of the withdrawal period when the milk can be given to the dairies for processing, samples were taken from the same animals, classified into three subgroups (C1 for penicillin and streptomycin; C2 for oxytetracycline; and C3 for enrofloxacin).It should be noted that most of the milk from these farms is supplied to local dairies, while a smaller portion is reserved by the producers' families for making artisan cheeses for their own consumption.In total, 500 samples of milk were analyzed, originating from the above groups of ewes.Milk samples were collected following a procedure that included cleaning the udder with lukewarm water, discarding the first three streams of milk, and disinfecting the teat ends with an alcohol solution (70% absolute alcohol).All samples were immediately transported to the laboratory within one hour of reception and kept at 4 • C for a maximum of 24 h before analysis.

Methodology: Isolation and Phenotypic Characterization of Escherichia coli
Each sample underwent thorough mixing in its sampling vial via 20-25 inversion movements.Subsequently, within 3 min of agitation, 10 mL of the sample was drawn using a sterile pipette and transferred to a dilution bottle containing 90 mL of homogenization liquid (Buffered Peptone Water, BPW, BioMerieux, Marcy l'Etoile, France), resulting in a 1:10 dilution.Decimal dilutions were subsequently carried out up to a dilution factor of 10 −8 , using Ringer's solution.A variety of selective and differential plating media has been used for the recovery of E coli isolates.The rationale for adopting this approach was to support the growth of both starved and unstarved E. coli cells.The following commercial media were used: Violet Red Bile Glucose Agar (VRBGA, Merck KGaA, Darmstadt, Germany), Endo Agar (Merck KGaA, Darmstadt, Germany), Ssorbitol McConkey agar containing 2.5 mg/L potassium tellurite and 0.05 mg/L cefixime (CT-SMAC, Merck KGaA, Darmstadt, Germany), CT-SMAC additionally supplemented with 50 mg/L 5-Brom-4-chlor-indolyl-beta-D-glucuronid cyclohexylammonium-salt (Merck KGaA, Darmstadt, Germany), CHROMagar™ E.coli, EC O157:H7 ChromoSelect Agar, modified and supplemented with 0.25 mL of 1% potassium tellurite solution/L (Merck KGaA, Darmstadt, Germany), CHROMagar™ O157 (CHROMagar Microbiology, Paris, France), supplemented with 2.5 mg/L tellurite and 2.5 mg/L cefsulodin, CHROMagar™ STEC (with 50 mg/L STEC supplement), and CHROMagar STEC supplemented with 1 µg/mL cefotaxime.The inoculated different solid media were incubated aerobically at 37 • C for 24-32 h.
Additionally, apart from employing standard procedures with chromogenic agar media, as described above, an enrichment protocol was implemented that consisted of two steps.Initially, 25 mL of homogenized sample is transferred to a sterile container containing 225 mL of modified Trypticase Soy Broth with the addition of 20 mg/L novobiocin (mTSB, Merck KGaA, Darmstadt, Germany).After 24 h of incubation at 37 • C, the second step involves the after-enrichment process, where a fraction of 100 µL of the 24 h TCS culture is inoculated onto sorbitol MacConkey agar (CT-SMAC) and supplemented with enrichment agents cefixime (0.05 mg/L) and potassium tellurite (2.5 mg/L), CHROMagar™ STEC (with 50 mg/L STEC supplement), and CHROMagar STEC supplemented with 1 µg/mL cefotaxime, followed by incubation at 37 • C for 24 h [37][38][39].
Bio-Typing: By Conventional Biochemical Tests and Platform VITEK 2 System After incubation, presumptive E. coli isolates were enumerated and subjected to phenotypic bio-typing procedures.These procedures involved a series of tests aimed at detecting specific metabolic characteristics, such as Gram staining, motility testing, cytochrome oxidase and catalase production, Triple Sugar Iron (TSI) Agar, and subsequently, the APIE 20E system (Biomerieux SA, Marcy l'Etoile, France) and the VITEK 2 automated microbiology system (bioMérieux, Marcy l'Etoile, France), all conducted according to the manufacturer's instructions.The latter system, employing the Vitek 2 ® ID-GN (Gram-negative, ID Card No. 21311) specific identification cards, performs comprehensive mapping of the biochemical properties of the isolates, alongside a reference strain E. coli ATCC 25922 utilized as a control.The confidence level was assessed at various levels, including excellent, very good, good, species, genus, low, and non-identification [40,41].

Sero-Typing: Serological Identification of E. coli Isolates
E. coli isolates were serotyped for the O157 and H7 antigens by slide agglutination using commercially available single SSI ® antisera E. coli and the corresponding pool systems against all recognized E. coli O antigens (O1 to O188) and H antigens (H1 to H56) (SSI Diagnostica, Hillerød, Denmark).
This was carried out in accordance with the manufacturer's instructions.Options such as using either boiled culture or live culture, or varying nutrient media and incubation temperatures were employed to induce the isolates to express the presence of the O antigen or their H flagellar.

Vero-Typing: Detection of Verocytotoxins Producing E. coli Serotypes
The presence of verocytotoxins (VT1 and VT2) produced by various E. coli serotypes was assessed using the Oxoid™ Reverse passive latex agglutination (VET-RPLA Toxin Detection Kit (Thermo Scientific™, Waltham, MA, USA).Each isolate was inoculated into CA-YE broth and incubated at 37 • C for 18-20 h with vigorous shaking (120-150 oscillations per minute).Following incubation, the culture was centrifuged at 4000 rpm for 20 min at 4 • C, and the supernatant was collected for the verocytotoxin assay.Each assay, conducted in triplicate, followed the manufacturer's instructions precisely [42].

Phenotypic Evaluation of Virulence Markers and Exoenzymes Production 2.5.1. Hemolysin Production
Hemolysin, also known as alpha-hemolysin, is a cytolytic toxin secreted by certain strains of E. coli.The evaluation of alpha-hemolysin production was conducted through the plate hemolysis method, utilizing blood agar plates.E. coli isolates were inoculated onto BD™ Columbia Agar with 5% Sheep Blood (Becton Dickinson, Heidelberg, Germany) and incubated aerobically overnight at 37 • C. Hemolysin production was indicated by the presence of a zone of complete clearance (halo) of erythrocytes surrounding the colony when observed against transmitted light.The extent of hemolysis (halo) was categorized as either positive (+) or negative (−) to denote hemolysin production and non-production, respectively [41,43].

Serum Resistance
Overnight cultures of E. coli isolates, cultivated at 37 • C on BD™ Columbia Agar with 5% Sheep Blood, were suspended in Hank's Balanced Salt Solution (HBS) containing 0.02% Phenol Red, filtered to 1000 mL through a membrane filter (0.45 µm pore size).Then, a bacterial suspension (0.05 mL) was incubated with 10% human serum (0.05 mL) at 37 • C for 180 min.Subsequently, 10 µL samples were extracted and spread onto 5% Sheep Blood agar plates, which were then incubated at 37 • C for 18 h to determine viable counts.Bacterial resistance to serum bactericidal activity was quantified as the percentage of surviving bacteria after 180 min of serum incubation relative to the initial count at time = 0. Bacteria were classified as serum-sensitive if the viable count decreased to 1% of the initial value (scored as negative), and as resistant if more than 90% of organisms survived after 180 min (scored as positive) [44].

Gelatinase Production Test
The gelatin hydrolysis test (gelatinase activity) was conducted utilizing Gelatin Agar Media (contains: 10 g/L gelatin, 5 g/L tryptone, 1 g/L glucose, 2.5 g/L yeast extract, and 20 g/L agar; final pH 7.2 ± 0.2).The plates were inoculated with the tested isolate and incubated at 37 • C for 24 h.Subsequently, the plates were flooded with a solution of mercuric chloride (150 g/L HgCl 2 in 20% v/v HCl).The appearance of opacity in the medium and the presence of transparent circles around the colonies were interpreted as positive indicators of gelatinase production [43,45].

Phenotypic Assessment Detection of Lipase, Caseinase, Lecithinase, and Amylase
The isolates were screened to assess the secretion of various virulence markers, including lipase, caseinase, lecithinase, and amylase enzymes, using a culture-based agar plate assay.These virulence factors, when secreted, are known to play roles in host tissue invasion through direct damage, complement system inactivation, cytotoxicity, and other mechanisms.Overnight cultures of each isolate were activated and streaked onto agar media containing specific substrates suitable for each enzyme activity.

-
The lipolytic activity (lipase production) of E. coli isolates was assessed using Tween 80-agar media (composed of 15 mL/L Tween 80, 5 g/L tryptone, 2.5 g/L yeast extract, 5 g/L NaCl, and 20 g/L agar; pH 7.2 ± 0.2).The plates were spot inoculated with the isolates and incubated aerobically for 24-48 h at 37 • C. The presence of clear zones (halos) around the colonies, following staining with methyl red solution (0.2 g/L methyl red in 95% ethanol), indicated the presence of lipolytic activity (lipase production) [46].- The production of caseinase from the isolates was detected using Skimmed Milk Agar (SMA, HiMedia Laboratories GmbH, Modautal, Germany).The isolates were streaked onto SMA plates and then incubated aerobically for 24 h at 37 ).These cultures were incubated again overnight at 37 • C. Subsequently, they were further diluted 1:100 in the same supplemented M9 medium, and 150 µL samples were dispensed in triplicate into a sterile 96-well microtiter plate (Costar ® 3370, Corning, NY, USA).E. coli ATCC 25922 served as a positive control for each experiment, while the M9 medium without any bacteria functioned as the negative control.After a 24 h incubation period at 30 • C, the wells were thoroughly washed three times with distilled water, dried at 37 • C for 30 min, and then stained with 0.1% (w/v) crystal violet for 2 min.Following another set of three washes and drying, the crystal violet was dissolved in 150 µL of an 80:20 (v/v) ethanol/acetone solution.Absorbance was measured at 595 nm and the biofilm production was categorized as none, weak, moderate, or strong based on previously published criteria.The optical density of the negative control (ODc) served as a baseline.Isolates were classified as non-biofilm producers if their optical density was less than or equal to the ODc (−), weak biofilm producers if their optical density was above the ODc but less than or equal to twice the ODc (+), moderate producers if their optical density was more than twice but less than or equal to four times the ODc (++), and strong biofilm producers if their optical density exceeded four times the ODc (+++).This procedure was repeated three times for accuracy [49,50].
Subsequently, 50 µL of each antibiotic solution was added to individual wells of a 96-well microplate.Then, 50 µL of the bacterial suspension (10 6 -10 7 CFU/mL) was dispensed into each well, where the various dilutions of antibiotics had been placed.The microplate was incubated aerobically at 37 • C for 24-32 h.The Minimum Inhibitory Concentration (MIC) was recorded as the lowest concentration of the antibiotic that halted the growth of the tested isolates.The results were categorized as sensitive, intermediate, or resistant based on the standardized table provided by NCCLS 2000, CLSI 2018, and EUCAST 2013 [51][52][53].E. coli ATCC 25922 and Staphylococcus aureus ATCC 6538 were used as quality controls.The experiment was repeated three times.Multidrug-resistant (MDR) E. coli was defined as bacteria that were resistant to at least three different classes of antimicrobials.

Detection of Extended-Spectrum Beta-Lactamase (ESBL) Producers Phenotype in E. coli Isolates
All isolates were assessed for extended-spectrum beta-lactamase (ESBL) production according to the criteria established by the CA-SFM/EUCAST [54,55], using the Combined Disc Synergy Method with MAST IDTM ESβL Detection Discs (Mast Group, Bootle, UK).The MAST Extended IDTM ESβL Detection Discs consist of discs containing cefpodoxime 30 µg, cefpodoxime 30 µg with clavulanic acid 10 µg, ceftazidime 30 µg, ceftazidime 30 µg with clavulanic acid 10 µg, cefotaxime 30 µg, and cefotaxime 30 µg with clavulanic acid 10 µg.A suspension of the test isolates in distilled water, equivalent in density to a McFarland 0.5 opacity standard, was prepared.This suspension was evenly spread across the surface of Mueller-Hinton agar plates (Merck KGaA, Darmstadt, Germany), using a sterile swab.One of each MAST IDTM ESβL Detection Disc was then placed onto the inoculated medium using sterile forceps, ensuring even spacing.The plates were aerobically incubated at 35-37 • C for 18-20 h.Any observed zones of inhibition were measured and recorded.The diameter of the inhibition zones for cefpodoxime, ceftazidime, and cefotaxime was compared to those of the respective combination discs containing clavulanic acid.An increase in zone diameter of ≥5 mm in the presence of clavulanic acid from any or all sets of MAST IDTM ESβL Detection Discs indicates the presence of ESBL in the tested E. coli isolates [56][57][58].

Statistical Analysis
Odds Ratio and a chi-square test of independence were used for a statistical significance level of 0.05.The Fisher exact test was employed to assess whether there was a non-uniform distribution of each serotype across various clinical and environmental sources.Prevalence rates and resistant phenotypes of the isolates were compared among the groups.Statistical analysis was conducted using SPSS software (version 20.0; SPSS, Chicago, IL, USA), with statistical significance set at a p-value below 0.05.

Results
Clarification: We would like to clarify that, for the sake of simplicity and economy, the subgroups (B1, B2, and B3) within group B are not mentioned in some tables.Instead, only the central subgroup, referred to as B, is listed.These animals exhibited clinical mastitis but had not yet started treatment, so they were considered as one group (B).Subsequently, three subgroups with different therapeutic approaches were created depending on the drug of choice (B1, B2, and B3).These three groups correspond to the three subgroups of the main group C (C1, C2 and C3), respectively.
Table 2 illustrates the distribution of the different types of E. coli isolated from raw milk among different groups and subgroups.The occurrence of the microorganism (total) in the milk from the healthy udders (group A, 5%) is significantly lower compared to milk from udders affected by mastitis (group B, 29.50%), which is very high, indicating a substantial difference (OR = 12.32, CI 95% 4.73-32.08,p < 0.001).Likewise, the prevalence of E. coli in the milk from animals treated with antibiotics (total of groups C1 + C2 + C3 = 17.00%) is lower than in the mastitis group B (OR = 2.29, CI 95% 1.39-3.81,p = 0.0013), but increased with respect to the controls (OR = 5.36, CI 95% 2.01-14.26,p = 0.0008).E. coli O157:H7 sorbitol (−), absent in the controls, shows a similar prevalence rate of 1.00-2.00% in all other groups.E. coli O157:H7 sorbitol (+) is absent in the controls as well as in the C3 group but present in all others with prevalence rates varying from 4% in group C2, to 0.50% in group C1, and to 0.50% in group B (clinical mastitis).The non-O157:H7 E. coli isolation rate was minimal in the control group (5%), excessively high in the mastitis group (27.50%), and reached intermediate values in the treatment groups: 8.00% in C1, 14.00% in C2, and 26.00% in C3 (Table 2).As for the relationship between the isolation frequencies of E coli O157:H7 strains [both sorbitol (-) and sorbitol (+)] with the isolation frequencies of non-O157:H7 E. coli strains, their distribution does not differ significantly among the different groups (p = 0.8172).
Tables 3 and 4 display the metabolic and virulence characteristics of the isolated strains of non-O157 E. coli, as well as the sorbitol-negative (−) and sorbitol-positive (+) E. coli O157 strains, respectively.There was a diversity of these traits among the different strains as expected but it could not be associated with the groups.However, the strains of E. coli O157 H7 sorbitol (−) showed a higher uniformity since all of them were amylase and lecithinase producers, reacted positively to the gelatinase test, had a positive serum resistance reaction, and all produced the serotoxin (STX) 2, while none of them was a lipase producer.*: Production or detection, accordingly; **: Enumeration of E. coli isolate and its corresponding experimental group designation.The interpretation of sampled groups, including the sources of milk samples, is presented in Figure 1.
Table 5 shows the resistance to antibiotics of the isolated E. coli strains in the different groups.The prevalence of the susceptible and resistant strains changed significantly during and after the treatments, from the B to the C groups.Spearman's rank test revealed a statistically significant monotonicity (r values > 0.80, p < 0.001) among the sensitive and resistant strains in each group.However, in every group, the antibiotic substance used for the treatment of mastitis led to an increased prevalence of resistant strains and a reduced prevalence of sensitive ones.Table 5 also displays the distribution of extended-spectrum β-lactamase (ESBL) phenotype production, which did not display a consistent pattern.For example, in groups C1 and C2, the majority of strains did not demonstrate ESBL activity, whereas the opposite trend was observed in group C3.Table 6 shows the resistance of the nine E. coli O157:H7 strains classified by group and by their reaction to sorbitol.All strains were found to be resistant to amoxicillin, and all were found to be sensitive to sulfonamides.
In Table 7 the capacity of biofilm formation of the isolated E. coli strains is presented.Most of these strains (52.17%) had the capacity to form biofilm.Of the 48 strains able to form a biofilm, 27 of them (56.25%)possessed this property in high intensity.In total, 38 out of 92 strains (Figure 2) were classified as multi-resistant (41.30%), 18 of which were isolated from the B group (47.36% of the MDRs).This finding implies that the Table 8 shows the number of non-O157:H7 E. coli strains, isolated from the different groups, representing many potentially pathogenic serotypes.For instance, all the socalled "big six" were present in the B groups and remained present-although with reduced prevalence-in the C groups.Mastitis significantly increased the prevalence of the "O" serotype strains (OR = 8.43, CI 95% 2.95-24.87,p < 0.0001) with respect to the controls.The treatments decreased this prevalence in the C group (OR = 0.44, CI 95% 0.27-0.74,p = 0.002) with respect to group B, but not with respect to the controls (OR = 3.75, CI 95% 1.27-11.02,p = 0.0165).The distribution of the "O" serotype strains varied significantly among the C subgroups (χ 2 = 9.337, df = 2, p = 0.0092).
Table 8.The number of non-O157:H7 E. coli strains, isolated from the different experimental groups.

Group
Non-O157:  Table 8 shows the number of non-O157:H7 E. coli strains, isolated from the different groups, representing many potentially pathogenic serotypes.For instance, all the so-called "big six" were present in the B groups and remained present-although with reduced prevalence-in the C groups.Mastitis significantly increased the prevalence of the "O" serotype strains (OR = 8.43, CI 95% 2.95-24.87,p < 0.0001) with respect to the controls.The treatments decreased this prevalence in the C group (OR = 0.44, CI 95% 0.27-0.74,p = 0.002) with respect to group B, but not with respect to the controls (OR = 3.75, CI 95% 1.27-11.02,p = 0.0165).The distribution of the "O" serotype strains varied significantly among the C subgroups (χ 2 = 9.337, df = 2, p = 0.0092).
Table 8.The number of non-O157:H7 E. coli strains, isolated from the different experimental groups.

Discussion
Milk is the foundational raw material of the dairy food chain, and its quality is crucial for the safety of dairy products.As a rich and complex nutritional matrix, milk provides an excellent medium for nearly all microorganisms.The health of the udder and the production site of milk significantly impact its hygienic status, especially in cases of subclinical or clinical mastitis.When mastitis occurs, treatment with antibiotics typically follows, curing the udder, and after the withdrawal period, the milk can be safely consumed again.This research aims to study the prevalence of pathogenic E. coli strains on the first day after the end of the withdrawal period when the milk re-enters the food chain.
Tables 2 and 8 display the prevalence of pathogenic E. coli strains across different experimental groups.Mastitis groups (B groups) notably harbor a high abundance of such strains, which, although significantly reduced, persist in milk after the withdrawal period (C groups).In group A, originating from healthy udders, the prevalence of E. coli isolates is 4%, a figure relatively lower or comparable to that reported by other researchers [24,[59][60][61].While some authors have reported higher prevalence rates (Ombarak et [62,63].It is important to note that E. coli is not commonly regarded as a major causative agent for mastitis in ewes [64]. Pathogenic E. coli O157 was not found in group A (healthy udders), consistent with findings from other studies by researchers such as Sancak et [2,31,65].It was detected in group B (4 strains, 2.67% of the samples) and maintained a similar occurrence (2-4%) in the C groups despite antibiotic treatment.Additionally, other pathogenic non-O157 E. coli isolates were isolated.The "big six" serotypes (O26, O45, O103, O111, O121, O145), known for their ability to cause severe illness, were present in all groups [66,67], along with other serotypes associated with illness such as O104 and O152 [68][69][70].
The classification of "O" serotypes is pivotal in epidemiology for linking cases to their source.Currently, a total of 176 "O" antigen structures of E. coli have been identified [71], and the conventional serotyping method, which involves agglutination specific to each serotype, continues to be utilized [72].In this study, 15 different "O" serotypes were isolated, 10 of which are considered hazardous.Since our samples were directly obtained by milking the udder and the microorganisms showed a low occurrence rate, it suggests that the infected udder is the source of these serotypes.When mastitis occurs, the bacterial populations of the normal microflora are significantly disrupted, allowing various commensals like E. coli to enter the ecosystem.Although most serotype prevalence rates decreased after treatment, not all did.For instance, in subgroup B3, three strains of O111 were isolated, while in subgroup C3, the number increased to four.This indicates contamination of environmental origin during the withdrawal period.The duration of the withdrawal period depends on the antibiotic class, and while it ensures the absence of antibiotic metabolites in milk, it does not guarantee the return of bacterial microflora to its normal equilibrium.Consequently, various opportunistic pathogens and commensals may still be present in the milk consumed by humans.
The virulence of E. coli relies on various factors including adhesion, biofilm formation, nutrient acquisition, competition with other bacteria, toxin production, and evasion of host defense mechanisms [73].Toxins such as cytotoxic necrotizing factor 1, autotransporter toxins, and alpha-hemolysin enhance E. coli virulence by targeting the cell's structure, metabolism, or cytoplasmic membrane [74].Each pathotype exhibits a distinct virulence profile determined by gene clusters located in the chromosome or mobile elements encoding specific proteins [75].Table 4 illustrates the virulence traits of E. coli O157 strains.A comparison between sorbitol-negative and sorbitol-positive strains suggests that sorbitolpositive strains are slightly more virulent.In this category, all strains produce both toxins, and more strains demonstrate biofilm capacity and exhibit ESBL resistance compared to sorbitol-negative strains.Table 3 depicts the virulence traits of non-O157 E. coli isolates, indicating they are generally less virulent than O157 isolates.There are notable differences in virulent metabolic traits between O157 and non-O157 isolates.All O157 isolates, regardless of sorbitol reaction, produce amylase and lecithinase, test positive for gelatinase, and exhibit serum resistance.However, among non-O157 isolates, the presence of these traits varies among groups.It is important to note that not all non-O157 isolates are less pathogenic, as many possess all aforementioned virulence traits.
Biofilm formation is a significant virulent trait in discussions concerning E. coli, particularly due to the "restaurant hypothesis".According to this theory, commensal E. coli colonizes the intestine as part of a biofilm microbiome, absorbing essential nutrients through cross-feeding from anaerobes that decompose polysaccharides.Interestingly, they also exhibit antagonistic behavior against individual planktonic E. coli cells [76][77][78].In our study (Table 7), 52.17% of the strains, regardless of their "O" classification, demonstrated biofilm-forming abilities, with almost half of these strains exhibiting this capacity to the highest degree.The survival of such strains in the C groups (46.74%) can be attributed to the increased protection provided by biofilms, even against antibiotics [79][80][81][82].
Mastitis treatment in our study was administered by professional veterinarians.Empirical selection of antibiotics at the farm level was the norm, meaning all animals affected by mastitis received the same antibiotic based on the farm's therapeutic history.Antibiotic susceptibility testing or other laboratory diagnostic methods were often avoided due to increased treatment costs, leading veterinarians to rotate antibiotics frequently to prevent the emergence of resistant strains.Farmers and farm personnel were experienced in recognizing clinical mastitis early and promptly calling veterinarians.They were also well informed about the importance of withdrawal periods, ensuring compliance to prevent milk rejection.
Our findings (Table 5) demonstrate varying resistance frequencies depending on the antibiotic class used, an expected observation, along with some unexpected discoveries.The twelve antibacterial substances utilized in this study belonged to five distinct classes based on their structure and mode of action.Group C1, receiving a combination of procainic benzylpenicillin and dihydrostreptomycin, showed less impact on β-lactam resistance, with increased resistance observed in strains against amoxicillin, which is structurally and functionally similar to procainic benzylpenicillin.Resistance to Cephalothin (CEP), a thirdgeneration β-lactam, was less prevalent, while resistance to cefotaxime, a fourth-generation β-lactam, remained unaffected.It appears that as β-lactams evolve, resistance mechanisms that were effective against older members of the family have a diminishing or negligible impact on newer members.The increased resistance frequencies to both streptomycin and chloramphenicol may be attributed to a mechanism of cross-resistance, as these antibiotics have different structures but similar modes of action (interfering with the 70s ribosomal subunit).Table 5 also highlights the capacity of isolated E. coli strains to produce ESBL, with 43.47% of strains demonstrating this phenotype in vitro.However, the expression of this phenotype in vivo depends on other genetic and environmental parameters, which were apparently not favorable in our study, as evidenced by the relatively lower resistance frequencies against Cephalothin (CEP) and cefotaxime (CEF).ESBL genes are known to circulate in the environment, with some studies suggesting sheep as reservoirs for these genes [83].A meta-analytical study by Bezabih et al. (2020) concluded that human intestinal carriage of ESBL E. coli globally increased by a factor of eight [84].Our results align with those of Obaidat and Gharaibeh (2022) [85].
In group C2, animals were administered oxytetracycline.Interestingly, while the resistance frequency to tetracycline decreased-a paradoxical finding-the resistance frequency to gentamicin increased, while resistance to amikacin remained unaffected.These three substances share the same mode of action (binding to the 30s subunit of the ribosome) but have different structures.The increased resistance to gentamicin may be attributed to a cross-resistant effect, and a similar explanation could apply to amikacin, a newer medication that might be less susceptible to standard mechanisms of resistance.It is possible that strains resistant to oxytetracycline were eliminated by other mechanisms, such as antagonism with other bacteria, preventing their survival and subsequent counting.
Finally, animals in experimental group C3 were treated with enrofloxacin, resulting in a significant increase in resistance frequency to ciprofloxacin.Interestingly, the resistance rate to nalidixic acid, which shares the same mode of action with quinolones by inhibiting the gyrase, decreased.This phenomenon could be attributed to the possibility, similar to that seen with oxytetracycline, that resistant bacteria were unable to survive due to antagonism with other microorganisms.
All isolates demonstrated sensitivity to the combination of trimethoprim/sulfamethoxazole (TRI/SU) and sulfathiazole (SUF), likely because neither of these substances nor any structurally or functionally related antibiotics (which interfere with the metabolism of folic acid) were included in the therapeutic protocols of local veterinarians.
A total of 41.30% of the E. coli isolates were classified as multidrug-resistant (MDR) as they exhibited resistance to more than two classes of antibiotics.It is concerning that 20 out of 38 MDR strains (52.63%) were isolated from the C groups.Among O157 E. coli isolates, all four from group B were multidrug-resistant (Figure 2), while three out of six strains from the C groups also demonstrated multidrug resistance.Interestingly, the two strains not classified as MDR were sorbitol-positive.
While molecular methods offer greater accuracy and precision in microbial characterization, conventional methods remain valid and widely used for classifying E. coli isolates [86].Bacterial identification using selective media is cost-effective and less complex, although it lacks the specificity and sensitivity of molecular biology techniques.Consequently, selective culture media are not highly reliable for identifying bacteria from diverse environments, and results must be verified using more accurate methods.Despite this limitation, selective culture media serve as a robust screening tool for preliminary bacteria selection before further molecular analyses.

1.
Udders affected by clinical mastitis harbor a diverse range of pathogenic E. coli "O" serotypes, including O157, at a higher prevalence compared to healthy and treated udders.2.
These isolates exhibit virulent traits, such as toxin production, hemolysin production, antibiotic resistance, ESBL activity, and biofilm formation.

3.
Antibiotic resistance is influenced by the therapeutic protocol employed, particularly the choice of antibiotic.4.
Even after the end of the withdrawal period, some strains of the E. coli "O" serotype remain in the milk, thus posing a serious threat to public health if this milk is consumed raw or if it enters the dairy process without pasteurization.

5.
The aforementioned threat is exacerbated by the relatively high prevalence of antibioticresistant isolates.

Figure 1 .
Figure 1.Interpretation of sampled groups-sources of testing milk samples.

Figure 1 .
Figure 1.Interpretation of sampled groups-sources of testing milk samples.

Figure 2 .
Figure 2. Mult-resistant strains of E. coli, isolated from studied experimental groups.Clarification: MDR is multidrug resistance, meaning resistance in more than 3 classes of antimicrobial agents; Non-MDR means resistance in less than 3 classes of antimicrobial agents.The experimental groups of animals are displayed on the horizontal axis.The interpretation of sampled groups, including the sources of milk samples, is presented in Figure 1.

Figure 2 .
Figure 2. Mult-resistant strains of E. coli, isolated from studied experimental groups.Clarification: MDR is multidrug resistance, meaning resistance in more than 3 classes of antimicrobial agents; Non-MDR means resistance in less than 3 classes of antimicrobial agents.The experimental groups of animals are displayed on the horizontal axis.The interpretation of sampled groups, including the sources of milk samples, is presented in Figure 1.

Table 1 .
Antimicrobial substances and therapeutic protocols in the different experimental groups.a BW: body weight; b qd: every day; c IM: intramuscular; d SC: subcutaneous.
MgSO 4 , 2 µM FeCl 3 , 8 µM CaCl 2 , and 16 µM MnCl 2 [47,48]e presence of transparent zones around the colonies indicated caseinase production[42].-Toassesslecithinaseactivity, isolates were cultured on an Egg Yolk Agar base (HiMedia Laboratories GmbH, Modautal, Germany) supplemented with 2.5% yolk.Incubation was conducted for 18-48 h initially, with further incubation for up to 7 days to account for potential delayed lipase activity.The plates were then examined for the presence of opaque zones around the colonies to confirm lecithinase production [42].-Foramylolyticenzymes, the E. coli isolates were streaked onto Starch Agar media plates containing 10 g/L soluble starch, 5 g/L tryptone, 3 g/L yeast extract, and 20 g/L agar at pH 7.2 ± 0.2.The plates were then incubated aerobically for 24-48 h at 37 • C. Following incubation, the plates were flooded with potassium iodide solution and allowed to react for 10 min.The presence of transparent zones surrounding the colonies indicated amylase production[47,48].

Table 2 .
Number of isolated E. coli strains in different experimental groups and total viable counts in log CFU/mL ± SD.
*: The interpretation of sampled groups, including the sources of milk samples, is presented in Figure1; **: number of animals (and samples) in the group; ***: number of strains isolated in the group.

Table 3 .
Metabolic and virulent traits of the non-O157:H7 E. coli strains, isolated in the different groups.
*: Production or detection; **: The interpretation of sampled groups, including the sources of milk samples, is presented in Figure1; ***: number of strains isolated in the group.

Table 4 .
Metabolic and virulent traits of the E. coli O157:H7 isolates from the different animal groups studied.

Table 5 .
The distribution of antimicrobial resistance phenotypes of the isolated E. coli strains in the different groups by mode of action of the tested antibacterial substances.

Production of extended spectrumβ-lactamase-ESBLs phenotypes
a : number of E. coli strains isolated in the experimental group.The interpretation of sampled groups, including the sources of milk samples, is presented in Figure1; b : S, I, R for susceptible, intermediate, and resistant

Table 7 .
Biofilm activity of the isolated E. coli strains in the different experimental groups.