Sixteen autochthonous Streptococcus spp. isolates originating from two batches of sheep milk of native Epirus breeds before (KFM isolates) and after (KTM isolates) thermization (65^oC; 30 s), and previously identified at the species level by 16S ribosomal RNA (rRNA) gene sequencing [23], were selected for further study (Table 1).

Two clinical Enterococcus strains were used as reference controls in the AR tests: Enterococcus faecalis ATCC 29212^TM [25], a quality control strain defined by the Clinical and Laboratory Standards Institute (CLSI) [26], and Enterococcus faecium 315VR, a virulent, MDR human isolate primarily serving as a vancomycin-resistant control strain [27]. Additionally, two basic NSC strains, Streptococcus thermophilus ST1 and the wild, nisin A-producing (NisA+) Lactococcus lactis subsp. cremoris M78 [27–29] were included in the present biochemical assays, as specified in the Results.

All sheep milk isolates and the control strains were resuscitated by transferring 0.1 mL of frozen (–30^oC) working stock cultures with 20% (w/v) glycerol, in 10 mL MRS broth (Neogen Culture Media, Lab M, Heywood, UK), incubated at 30^oC for 24 h, and then subcultured twice, as above. Particularly the thermophilic Streptococcus isolates of the SESB complex [20] were also resuscitated and subcultured in 10 mL M17 broth (Merck, Darmstadt, Germany), incubated at 37^oC for 24 h to ensure their optimal recovery and growth. The second subculture of all isolates was streaked for growth on MRS agar (Neogen) or M17 agar (Biolife, Italiana, S.r.l., Milano, Italy) at 30^oC or 37^oC for 48–72 h, to check and ensure purity by transferring one single colony to new 10 mL MRS or M17 tubes, incubated as above, for use in the experiments.

All autochthonous sheep milk isolates were rechecked for their Gram-positive and catalase-negative reactions and basic phenotypic traits: microscopic appearance, CO₂ production from glucose, NH₃ production from arginine, growth at 10^oC and 45^oC, growth in 4% and 6.5% NaCl, and growth on Kanamycin Aesculin Azide (KAA) agar (Neogen). Next, all isolates were tested for acid production from 13 basic (key) sugars, L-arabinose, cellobiose, galactose, lactose, maltose, mannitol, melibiose, raffinose, ribose, sorbitol, sucrose, trehalose, and xylose (Sigma-Aldrich Chemie GmbH, Steinheim, Germany), in pre-sterilized 96-well mini-plates (Sarstedt AG & Co. KG, Nümbrecht, Germany) to identify them biochemically at the species, in comparison with their 16S rRNA species identification (Table 1). The mini-plates were incubated at 30^oC for 72 h or at 37^oC for 48 h for the mesophilic and thermophilic isolates, respectively; the positive sugar fermentation readings were based on the color change of the bromocresol purple indicator to yellow due to acid production. The biochemical discrimination and identification of animal Streptococcus spp. was based on the sugar fermentation patterns tabulated in Bergey’s Manual of Systematic Bacteriology [11].

Additionally, all 16 Streptococcus sheep milk isolates (Table 1) were identified by the API 20 STREP biochemical method, according to the manufacturer’s instructions (BioMerieux, Marcy l’Etoile, France). This commercialized id-kit detects for 10 key taxonomic sugar fermentation reactions and 10 key enzymatic activities within the genus Streptococcus and related coccoid LAB genera, mainly Enterococcus; a positive or negative β-hemolysis reaction of each strain is required as an additional external test to generate its 7-digit API 20 STREP identification code. According to Sioziou et al. [23] and Samelis et al. [6], all streptococcal isolates in Table 1 were not β-hemolytic, but were α-hemolytic. Therefore, the hemolysis test was not repeated during this study.

However, in the course of this study, safety evaluation of Streptococcus spp. by biochemical tests was extended to biogenic amine (BA) formation, an undesirable trait for the selection of autochthonous starter or adjunct LAB strains. Therefore, all sheep milk isolates were tested for their in vitro ability to produce histamine and tyramine by using the improved BA screening method of Bover-Cid and Holzapfel [30]. BA production was recorded by the color change of the improved BA broth from yellowish to blue or purple, using E. faecalis ATCC 29212^TM and E. faecium 315VR as tyramine-positive controls, as described by Tsanasidou et al. [27].

All Streptococcus sheep milk isolates were tested for susceptibility to eight common antibiotics, according to the CLSI [26] protocol using the Kirby-Bauer disk diffusion method (μg per disk): ampicillin (AMP; 10), chloramphenicol (CHL; 30), ciprofloxacin (CIP; 5), erythromycin (ERY; 15), gentamicin (GEN; 10), penicillin G (PEN; 10 units per disk), tetracycline (TET; 30), and vancomycin (VAN; 30), as described by Tsanasidou et al. [27]. For comparative purposes, the AR testing of the present isolates was conducted together with all Enterococcus spp. strain biotypes co-isolated from the same RM/TM samples using the same agar media and reagents [24]. Briefly, commercial antibiotic disks (BioMaxima SA, Lublin, Poland) were used. Two M17 agar plates were spread-inoculated with 100 μL of fresh (24-h) M17 broth cultures of each tested isolate, the inocula were left to absorb for 5–10 min, and then 4 different antibiotic disks were placed over the cell lawn on each plate. All plates were incubated at 37^oC overnight. On the next morning, the diameter of the inhibition zones formed around each disk, including the diameter (6-mm) of the disk, as specified by CLSI [26], was measured with a micrometer (model D15, Mitutoyo, Kanagawa, Japan). The above process was repeated using a series of new fresh cultures, and the mean of the two independent replicate measurements of AR zone for each disk was calculated. M17 agar was used for the AR tests because streptococci and enterococci grow abundantly on this medium exclusively containing lactose, the primary sugar fermented by LAB in milk and dairy products, instead of glucose. However, the AR of some strains was further tested comparatively on Mueller-Hinton agar (Oxoid CM0337, Oxoid Ltd., Basingstoke, UK) medium [26] to account for potential media effects on the size and clarity of the inhibition zones. The results were interpreted according to the breakpoints recommended by CLSI [26] for streptococci and enterococci. As indicated in Sheep milk isolates, reference strains and culture conditions, the CLSI quality control strain E. faecalis ATCC 29212^TM (i.e., penicillin-resistant) and the MDR strain E. faecium 315VR served as positive controls.

All Streptococcus sheep milk isolates were further tested for their ability to ferment milk in pure culture in vitro compared with the NSC strains S. thermophilus ST1 and Lc. lactis M78. The acidifying activity rate was evaluated by measuring the pH declines (ΔpH) of inoculated (1% v/v) 10% reconstituted skimmed milk (RSM; Neogen) cultures in triplicate 10-mL tubes for each strain grown at 37°C for 6 h, shifted to 30°C for up to 24 h, and finally to 22°C for an additional 24 h, in simulation of the gradual temperature reductions of fresh curds during traditional cheese processing. The pH of all RSM cultures was measured by direct immersion of the glass electrode of a 3510 digital pH meter (Jenway, Essex, UK) in the milk. During incubation and before each pH measurement, the clotting strength and the uniformity of the curd, if any, were recorded macroscopically. Moreover, to check whether the temperature shifting from 37^oC (simulating the cheese milk curdling/fermentation step) to 22^oC (simulating the cheese ripening step) retarded the acidifying activity of the isolates, particularly the members of the thermophilic SESB complex (Table 1), additional RSM cultures inoculated as above were incubated at the optimal growth temperature for each species (i.e., 30^oC, 48 h for S. parauberis; 37^oC, 24 h for the SESB complex), and the final RSM pH values and clotting strength were recorded as above and compared. Finally, it should be noted that screening for the milk acidifying activity of the Streptococcus isolates during this study was limited to pH measurements; the titratable acidity for the strains potentially selected as natural starters should be measured in real cheese milks.

Fresh (37^oC; 24 h) MRS broth cultures of all sheep milk isolates were streaked on M17 and MRS agar plates. Following growth at 37^oC for 48 h, all plates were screened for direct antagonistic activity against the target strain Listeria monocytogenes no.10 by the simple agar overlay technique [23]. For the Streptococcus spp. isolates causing inhibition halos, bacteriocin activity was tested further by using well diffusion assays: filter-sterilized MRS culture (37^oC; 24 h) cell-free supernatants (CFS) were assayed directly or after adjustment to pH 6.2, heating at 100^oC for 5 min, and treatment with 1 mg/mL of proteinase K and trypsin (Sigma Aldrich Chemie GmbH, Steinheim, Germany) at 37^oC for 3 to 24 h, as described for enterococci by Sioziou et al. [23].

Finally, the two selected antilisterial strains S. lutetiensis KFM 55 and KFM 60 [23] were assayed for 19 constitutive enzymes in their cell biomass using the API ZYM semi-quantitative method, according to the manufacturer instructions (BioMerieux) and in direct comparison with the basic NSC strains S. thermophilus ST1 and Lc. lactis M78. Selection was made in view of their potential suitability for inclusion in NSCs or use as protective adjunct strains in future cheese making studies.

The basic phenotypic/biochemical differentiating characteristics of the autochthonous Streptococcus sheep milk isolates are presented in Table 2. First, it was noted that the nine thermophilic, arginine-negative Streptococcus sheep milk isolates molecularly assigned to the non-QPS species S. equinus, S. gallolyticus, and S. lutetiensis were not salt tolerant because all failed to grow in 6.5% salt (Table 2). Likewise, the mesophilic, arginine-positive S. parauberis isolates did not grow in 6.5% salt (biotype I) or grew with delay (biotype 2). Of note, all streptococcal isolates with cells forming long chains were S. equinus. This animal-associated species was subdominant in RM1 (KFM 5), but it prevailed in TM1 post-thermally (Table 1). In contrast, consistent with its mesophilic nature (Table 2), S. parauberis was not isolated from TM2, despite it prevailed in RM2 [23]. S. parauberis RM isolates formed two distinct biotypes, I and II, with the biotype I isolates failing to produce acid from cellobiose and sorbitol (Table 2). Notably, all three biotype II isolates (KFM 34, KFM 35, and KFM 42) matched one strain genotype in BLAST, too (Table 1), while all seven isolates shared the (key) sugar fermentation reactions with the species S. parauberis [11]. On the other hand, the basic sugar fermentation profiles of the single KFM 26 isolate from RM1 and both KFM 55 and KFM 60 isolates from RM2 were identical (Table 2), despite their 16S rRNA gene identification as two separate species of the SESB complex, S. gallolyticus and S. lutetiensis, respectively (Table 1). All differed from the RM/TM group genotyped as S. equinus (Table 1) by their inability to ferment cellobiose and melibiose (Table 2).

Therefore, to discriminate the present sheep milk isolates of the SESB complex at, and within, the species, it was necessary to use additional, more specific biochemical identification tools. For this purpose, all Streptococcus isolates were tested by the API 20 STREP method and their 7-digit id-codes generated according to the manufacturer’s instructions, were submitted to the Veterinary Laboratory, Food Hygiene Department, Ministry of Agriculture, Athens, for validating their identification using the apiweb^TM software; the results are presented in Table 3. Based on the recorded codes, only the S. gallolyticus KFM 26 strain genotype underwent an excellent identification; its previous 16S rRNA-based assignment to this species (Table 1) was extended to the subspecies pasteurianus [31] by the API 20 STREP method (Table 3). Additionally, a very good identification within the SESB complex as S. bovis biotype I/II was recorded for the two thermophilic isolates previously identified as S. lutetiensis (KFM 55) and/or S. lutetiensis/S. infantarius (KFM 60) by 16S rRNA gene sequencing (Table 1). This result was considered valid because the above two species were extracted and raised as novel species after a double sequential reclassification of the heterogeneous SESB complex [32, 33]. Specifically, the former S. infantarius subsp. coli [32] was raised to S. lutetiensis sp. nov. [33], the latter differing, among others, from S. infantarius subsp. infantarius in being β-glucosidase and esculin-positive [32]. Therefore, according to their positive esculin (β-glucosidase) hydrolysis reaction (Table 3), both KFM 55 and KFM 60 strains clearly belong to S. lutetiensis (formerly S. bovis biotype II.1) [33].

In contrast, the 16S rRNA identification of the KFM 5 isolate and thereby of the five TM isolates sharing this strain biotype as S. equinus (Table 1) was not consistent with their very good or good identification as Streptococcus pneumoniae by the API 20 STREP method (Table 3). Likewise, there was a discrepancy between the molecular identification of S. parauberis biotypes I and II by the 16S rRNA method using BLAST (Table 1) and their ‘acceptable’ or ‘uncertain’ identifications as Streptococcus agalactiae and S. uberis, respectively, using the API 20 STREP id-database (Table 3).

Lastly, regarding cheese milk safety, none of the non-hemolytic Streptococcus spp. sheep milk isolates biotyped in Tables 2 and 3 were found to produce histamine or tyramine in improved BA broth, unlike the two virulent control strains E. faecalis ATCC 29212^TM and E. faecium 315VR, which were confirmed to produce tyramine when tyrosine was added as a precursor in BA broth (data not shown).

The phenotypic AR profiles of the 16 Streptococcus spp. isolates originating from two sheep milk batches of native Epirus breeds before and after thermization are presented comparatively to the AR profiles of the two Enterococcus control strains in Table 4.

As anticipated, the vanA+ E. faecium 315VR strain was MDR whereas the CLSI quality control E. faecalis ATCC 29212^TM was resistant to penicillin but partially susceptible to most of the remaining antibiotics tested, including vancomycin (Table 4). Because AR is strain-dependent, every RM/TM isolate found resistant to at least one of the tested antibiotics is tabulated separately; otherwise, the remaining antibiotic susceptible or of intermediate sensitivity isolates within each species are combined by averaging the diameter of their mean inhibition zones.

Contrary to our anticipation, all Streptococcus isolates from sheep milk before or after thermization had restricted ARs (Table 4). Specifically, the RM2 S. parauberis biotype I isolates were susceptible to all antibiotics tested, except for KFM 41 that was resistant to ciprofloxacin and tetracycline, while all RM2 biotype II isolates were resistant only to tetracycline, and one of them (KFM 42) to ciprofloxacin similarly to KFM 41 above. S. equinus KFM 5 from RM1 was strongly resistant to tetracycline, too. However, from the five S. equinus isolates from TM1, two were close to the CLSI breakpoint (≤ 18 mm) and another two were of intermediate sensitivity to tetracycline (Table 4). None of the six S. equinus isolates were resistant to ciprofloxacin, but it is worthy of noting that the two tetracycline-resistant isolates (KTM 8, KTM 9) from TM1 were of intermediate sensitivity to ciprofloxacin as well. Conversely, S. gallolyticus subsp. pasteurianus and S. lutetiensis strain biotypes represented by the raw sheep milk isolates KFM 26 and KFM 55 = KFM 60, respectively, were fully resistant to ciprofloxacin, but susceptible or partially susceptible (i.e., KFM 26) to tetracycline and the remaining six antibiotics tested. None of the 16 Streptococcus sheep milk isolates of the S. uberis group or the SESB complex exhibited multidrug resistance (i.e., resistance to ≥ 3 antibiotic classes). Instead, all showed a remarkably high susceptibility to ampicillin, chloramphenicol, erythromycin, gentamycin, and mainly penicillin and vancomycin (Table 4).

The above AR profiles of the Streptococcus spp. isolates were confirmed for total eight representative isolates resistant to tetracycline and/or ciprofloxacin when tested on M17 agar plates (Table 4), which were also tested on Mueller-Hinton agar, the reference CLSI agar for disk diffusion AR testing [26]; the results are summarized in Table S1. Generally, the antibiotic inhibition zones on Mueller-Hinton agar (containing no glucose or lactose, but starch plus beef dehydrated infusion and casein hydrolysate) were similar, in some cases smaller, than those formed by the same strain on M17 agar; however, the AR profile of each representative strain was not critically altered. Most discrepancies were minor and noted with ciplofloxacin (Table S1). In fact, the most prominent and thus critical difference was the resistance of S. equinus KFM 5 to ciprofloxacin on Mueller-Hinton agar, whereas it was sensitive on the respective M17 agar plates (Table 4). To address this discrepancy, the AR testing on M17 agar was repeated particularly for this strain and confirmed that ciprofloxacin caused a large (22.9 mm) inhibitory zone on its cell lawn as all antibiotics (> 21 mm) did, except for tetracycline (14.8 mm) to which S. equinus KFM 5 was resistant on both agar media. Further tests are needed to clarify this medium-dependent AR discrepancy.

When inoculated (1% v/v) in heat-sterilized (121^oC; 5 min) RSM (initial pH 6.78) and grown at 30^oC for 48 h, the mesophilic S. parauberis biotype I and II isolates (Table 2) reduced the pH to 4.99 ± 0.02 (ΔpH 1.79) and 5.09 ± 0.04 (ΔpH 1.69), respectively; this slightly higher milk acidifying capacity of the biotype I isolates resulted in a firmer curd compared to the biotype II isolates which gave a softer unstable curd that was flowed by inverting the tube. It is worth noting that the RSM cultures of all S. parauberis isolates remained non-clotted (pH > 5.2) after the first 24 h, requiring at least a 42-h incubation period at 30^oC to curdle the milk. Conversely, the six thermophilic S. equinus isolates (Table 2) were the best acidifiers among the sheep milk streptococci in this study since their RSM culture pH fell to 4.80 ± 0.02 (ΔpH 1.98) and the skimmed milk was curdled firmly at 37^oC for 24 h. Both S. lutetiensis KFM 55 and KFM 60 strains were relatively less aciduric (pH 4.89 ± 0.09; ΔpH 1.89) and formed a softer curd than the S. equinus isolates in RSM. S. gallolyticus KFM 26 was the least acidifying strain (pH 5.34 ± 0.08; ΔpH 1.44) following growth in RSM at 37^oC for 24 h (data not tabulated).

From the above findings, it became evident that the gradual acidifying activity of the Streptococcus sheep milk isolates was primarily dependent on the species, less so on each strain within a species, and strongly on the growth temperature, as indicated by the kinetic acidification results presented separately for each isolate in Table 5. Indeed, the S. equinus isolates were the strongest, and the S. parauberis isolates were the weakest RSM acidifiers during the first 6 h at 37^oC, with the S. lutetiensis and S. gallolyticus isolates being of intermediate rapid acidification capacity. Surprisingly, based on the second-term ΔpH values during incubation, retardation of the acidifying activity of all S. parauberis isolates was similar (0.16 to 0.29 pH units) with that of the thermophilic SESB isolates (0.16 to 0.27 pH units) following shifting of the incubation temperature from 30^oC to 22^oC after the first 24 h (Table 5). Therefore, all S. parauberis failed to reduce the final RSM pH < 5.2 and curdle the milk when growth was completed at 22^oC, instead of 30^oC and above. Neither of the two S. lutetiensis strains curdled the RSM under the shifting incubation temperature conditions (Table 5), as both did so slowly at 37^oC for 24 h. Only S. equinus isolates managed firm RSM curds at 22^oC thanks to their superior milk acidification rate at 37^oC and then at 30^oC, while S. gallolyticus KFM 26 was the only raw sheep milk strain likely favored by the growth temperature shifting (Table 5).

Meanwhile, the primary starter S. thermophilus ST1 grew abundantly in RSM, managing a ΔpH value as high as 1.79 after 6 h at 37^oC, which increased further to 2.70 pH units after 24 h at 30^oC and resulted in strong curd formation early during incubation. The NisA+ costarter Lc. lactis M78 was a remarkably slower milk acidifier than ST1 but faster and stronger than all Streptococcus sheep milk isolates (Table 5).

All S. parauberis sheep milk isolates and mainly S. gallolyticus KFM 26 and both S. lutetiensis KFM 55 and KFM 60 strains inhibited growth of L. monocytogenes no. 10 on the MRS agar (initial pH 5.7 ± 0.1) overlays (i.e., all six S. equinus isolates were unable to grow in/on MRS media and thus could not be tested accordingly). However, none of the present S. parauberis, S. equinus, and S. gallolyticus strains could inhibit Listeria growth in the respective M17 agar (initial pH 7.2 ± 0.2; containing 0.5% lactose instead of 2% glucose) overlays. As previously noted [23], S. lutetiensis KFM 55 and KFM 60 were the only streptococcal raw sheep milk isolates that showed a clear bacteriocin-like antilisterial activity on M17 agar overlays (Figure 1A; the non-active KFM 26 strain is illustrated in Figure 1B for comparison). Next, the direct antilisterial activity of both S. lutetiensis strains was retained in their filter-sterilized acidic (pH 4.5–4.7) CFS after growth in MRS broth at 37^oC for 24 h (Figure 1C), as well as after the acidic CFS was neutralized at pH 6.2 and boiled at 100^oC for 5 min, but it was lost after treatment with proteolytic enzymes, proteinase K and trypsin (Figure 1D). Thus, strains KFM 55 and KFM 60 excreted a heat-resistant antilisterial peptide in MRS broth, most likely a class II bacteriocin. Further studies are needed to find out whether strains KFM 55 and KFM 60 possess one or more known or novel streptococcal bacteriocin gene/s in their genome and whether the active peptide/s can be produced in situ in sheep milk during cheese making or enrichment of the milk with glucose and possibly additional nutrients present in MRS is a prerequisite for bacteriocin formation by S. lutetiensis.

Based on the AR and biochemical identification results, and particularly on the API 20 STREP records closely relating most strain biotypes of sheep milk streptococci with the pathogenic species S. agalactiae, S. uberis, S. pneumoniae, and S. gallolyticus subsp. pasteurianus (Tables 2–4), only the two bacteriocin-producing S. lutetiensis KFM 55 and KFM 60 strains were further evaluated for their entire enzymatic activity profiles. Major species-dependent variations in the API ZYM profiles of the two S. lutetiensis and the two NSC strains, S. thermophilus ST1 and Lc. lactis M78 were noted (Table 6).

Specifically, both S. lutetiensis isolates displayed very strong α-galactosidase activity, which was absent in both NSC strains ST1 and M78, which, however, displayed strong β-galactosidase and α-glucosidase activity, respectively (Table 6). Actually, these were the most prominent API ZYM differences between the four strains as far as the crucial glycolytic enzyme system for milk fermentation is concerned. On the other hand, similarly to ST1 and M78, both S. lutetiensis strains displayed strong leucine arylamidase and acid phosphatase, and moderate napthol-AS-BI-phosphohydrolase activity. However, unlike M78, both S. lutetiensis strains lacked esterase (C4), esterase-lipase (C8), or valine and cystine arylamidase activities, which were also weak or absent in ST1. Moreover, both S. lutetiensis strains lacked α-chymotrypsin activity moderately possessed by M78, but not by ST1. Also, although both of them displayed a weak to moderate alkaline phosphatase activity in the API ZYM kit (Table 6), they should be considered negative as they were in the API 20 STREP id-kit (Table 3), which is more reliable for this particular enzymatic test [24]. Finally, no strain showed lipase (C14), trypsin, β-glucuronidase, β-glucosidase, N-acetyl-β-glucosaminidase, α-mannosidase, and α-fucosidase activities (Table 6).

Despite S. lutetiensis is a non-QPS species of the SESB complex that causes mastitis, selected aciduric or antagonistic strains of S. lutetiensis from artisanal cheeses lacking streptococcal virulence and AR genes have been proposed as starters or adjuncts. In alignment, the antibiotic susceptible (except ciprofloxacin) S. lutetiensis KFM 55 and KFM 60 strain biotypes from RM of native Epirus sheep breeds possess beneficial fermentative and enzymatic activities in vitro and antilisterial bacteriocin activity. The structural, either known or novel, bacteriocin gene/s or other antagonistic mechanism underlying the in vitro bacteriocin-mediated activity against L. monocytogenes has yet to be elucidated, along with a complete safety evaluation of these strains. Although both lack β-hemolytic activity and the vanA, vanB, gelE, hyl or cyl genes and are not MDR, the possibility of possessing other virulence genes associated with the SESB complex and the species S. lutetiensis in particular remains. Therefore, until their safety status is fully clarified, the inclusion of the phenotypically ciprofloxacin-resistant S. lutetiensis KFM 55 and KFM 60 strains in complex NSCs for commercial use in traditional Greek cheeses is hampered. Whole-genome sequencing and in situ cheese trials are required before recommending these strains for commercial NSC development.