Unless otherwise stated, media and anaerobic systems were purchased from Oxoid (Basingstoke, Hampshire, United Kingdom), while the chemicals were purchased from Sigma Aldrich (St. Louis, MO, United States). Primers and Sanger sequencing were provided by BMR Genomics (Padova, Italy), while the molecular biology reagents by Thermo Fisher Scientific (Waltham, MA, United States).

Type strains L. helveticus DSM 20075^T, L. fermentum DSM 20052^T, L. delbrueckii subsp. bulgaricus DSM 20081^T, L. delbrueckii subsp. lactis DSM 20073^T, and S. thermophilus DSM 20176^T served as reference in species attribution and were cultured according to the DSM guidelines. For cross-feeding experiments, the strains S. thermophilus RBC06 and C001.27, L. helveticus C001.15 and LBB04, and L. delbrueckii subsp. lactis T1104 and C5I3 were originally isolated from NWS using M17 and MRS media, as specified in Supplementary Figure 1. All strains were subsequently deposited in the culture collection of the Lactic Acid Bacteria and Yeast Biotechnology (LYB) Laboratory of the Department of Life Sciences (University of Modena and Reggio Emilia). Strains were routinely propagated under the same isolation conditions, as detailed in Supplementary Table 1.

NWS samples were collected in winter 2023 from two PR dairies located in the province of Reggio Emilia (Italy), both belonging to the Parmigiano Reggiano PDO Cheese Consortium. The samples, previously classified as type-H and type-D (Sola et al., 2022), were aseptically collected immediately after overnight whey incubation and stored under refrigerated conditions until analysis. All analyses were performed on the same day as sampling.

Total microbial counts (TMC) were determined using a Bürker chamber by counting cells in at least 10 randomly selected quadrants under a Nikon ECLIPSE 80i microscope at 400 × magnification. Cell viability in NWS samples was assessed using the LIVE/DEAD BacLight Bacterial Viability Kit (Thermo Fisher Scientific). Detailed protocols for TMC determination and viability assessment are reported in Cristofolini et al. (2025).

Titratable acidity (expressed in Soxhlet-Henkel degrees, °SH/50 mL), pH, and fermentative activity (defined as acidification rate and expressed as Δ°SH/50 mL) were determined following the procedures described by Cristofolini et al. (2025).

Bacterial cells were harvested from 2 mL of each NWS sample by centrifugation at 9,000 g for 10 min. After cell washing with TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0), total DNA was extracted using the DNeasy PowerSoil 96 Pro Kit (Qiagen, Hilden, Germany) on a QIAcube HT instrument (Qiagen, Hilden, Germany). The integrity and concentration of the extracted DNA were assessed by 1.5% agarose gel electrophoresis. Libraries were prepared and subjected to Illumina NovaSeq X platform by BMR Genomics (Padua, Italy). Specifically, libraries were prepared using the Illumina DNA Prep kit with IDT for Illumina UD Indexes, with adapter trimming performed using CTGTCTCTTATACACATCT, i7 adapters CA AGCAGAAGACGGCATACGAGAT[i7]GTCTCGTGGGCTCGG, and i5 adapters AATGATACGGCGACCACCGAGATCTACAC[i5] TCGTCGGCAGCGTC. Library quality was evaluated by measuring DNA concentration with a Qubit fluorometer (Thermo Fisher Scientific, Waltham, MA, United States) and fragment size distribution using a Bioanalyzer (Agilent Technologies, Santa Clara, CA, United States). Libraries were pooled by mass and sequenced on an Illumina NovaSeq X platform in paired-end 150 bp mode (2 × 150 bp). Raw sequencing reads in FASTQ format were preprocessed with fastp to remove adapters and low-quality bases, using a qualified Phred score of 20, an unqualified base limit of 30%, an average quality threshold of 25, and low-complexity filtering with a complexity threshold of 30 (Chen et al., 2018). Filtered reads were aligned with Bowtie (v.2.2.3) (Langmead et al., 2009) using default parameters against the SILVA v138 ribosomal RNA database (Quast et al., 2013) and the Homo sapiens GRCh38 genome to remove potential host contamination. Taxonomic profiling of microbial communities was then performed using MetaPhlAn v4 (June 2023 database) (Blanco-Míguez et al., 2023). Metataxonomic profile datasets generated for this study have been deposited in the NCBI GenBank database under the BioProject accession number PRJNA1367500.

Media used for culturomics analyses were listed in Table 1. MRS and M17 media were prepared according to the manufacturer’s instructions, with bacteriological agar added at a final concentration of 1.5% (w/v) when required. M17 and MRS media were supplemented with sterile skimmed whey (SSW; Reire, Reggio Emilia, Italy) at a final concentration of 3.5 g/L. A 5% (w/v) SSW stock solution was prepared as described by Fornasari et al. (2006). Yeast extract (YE) and tryptone (TRYPTO) were added to MRS at final concentrations of 1, 2, and 4% (w/v) before autoclaving. Lactalbumin and casein hydrolysates were prepared as 10% (w/v) stock solutions, sterilized, and added to MRS medium at 1, 2 and 4% (w/v), after autoclaving. L-cysteine (C) was added to MRS medium prior to autoclaving at a final concentration of 0.5 g/L, while formic acid (F) was aseptically added after autoclaving to MRS medium at a final concentration of 5 mM. Folic acid (Fa) was prepared as a 50 μg/mL stock solution in ddH₂O, filter-sterilized (0.2 μm), and added to MRS medium after autoclaving at a final concentration of 226 μM.

All media used for bacterial enumeration were supplemented with cycloheximide (50 μg/mL) to inhibit yeasts growth.

YPDA medium was prepared with 1% (w/v) yeast extract, 2% (w/v) peptone, 2% (w/v) glucose, and 1.5% (w/v) agar, to monitor generalist yeasts as contaminants in NWS. Chloramphenicol (50 μg/mL) was added to inhibit bacterial growth.

NWS samples were tenfold serially diluted in sterile saline solution (9 g/L NaCl) and plated by surface spreading using a sterile L-shaped spreader. Plates were incubated according to incubation conditions reported in Table 1. Specifically, lactobacilli were enumerated by incubating the plates at 42°C for 48–72 h under anaerobic conditions; streptococci were enumerated by incubating M17-SSW plates at 42°C for 72 h under aerobic conditions, as described by Fornasari et al. (2006); yeasts were enumerated by incubating YPDA plates at 42°C for 48 h under aerobic conditions. Viable cell counts were recorded as colony-forming units (CFU/mL) from plates containing 20–200 colonies and expressed as Log₁₀ CFU/mL, calculated as the mean of at least three replicates.

Individual colonies were randomly selected from plates containing 20–200 colonies and sub-cultured through three successive rounds of purification on the same isolation medium. The number of viable and cultivable isolates was monitored at each round of streaking and compared to the initial number of colonies selected from the primary plates to assess survival. Cultivability was expressed as the percentage of surviving colonies after three rounds of purification, according to Equation 1:

where N_d is the number of colonies that died during the three rounds of purification; N_v is the number of surviving colonies after three rounds of purification; N_i is the number of colonies initially isolated from the starting plate.

Axenic strains were examined for catalase activity, Gram-staining, and morphology, and subsequently stored at −80 °C in liquid medium supplemented with 25% (v/v) glycerol.

Genomic DNA was extracted by mechanical lysis from LAB isolates harvested at the early stationary phase, followed by organic solvent extraction, and quantified spectrophotometrically, as previously reported (Tagliazucchi et al., 2020). LAB isolates were identified by pentaplex PCR according to Cremonesi et al. (2011), with the exceptions that primer concentrations were reduced to 1 μmol/L and PCR conditions followed those described by Martini et al. (2024). When required, 16S amplified ribosomal DNA restriction analysis (16S-ARDRA) using the diagnostic endonucleases MseI and EcoRI and 16S rRNA gene sequencing were performed as previously described (Giraffa et al., 1998; Sola et al., 2022).

Lactobacillus delbrueckii subsp. lactis strains C5I3 and T1104 were grown on the corresponding isolation medium (Supplementary Table 1) until the late exponential phase and then inoculated into 5 mL of MRS, M17-SSW, C, CF, or CFFa media at a final cell suspension of app. 10⁵ CFU/mL. Cultures were incubated at 42°C for 72 h under anaerobic conditions, after which growth was monitored by measuring optical density at 600 nm (OD₆₀₀) using a microplate reader (Multiskan SkyHigh, Thermo Fisher Scientific, United States). Growth was expressed as percentage relative to the OD₆₀₀ values of the control condition (MRS medium), as reported by Fontana et al. (2019). All growth experiments were carried out in triplicate.

Cross-feeding experiments were conducted in milk using three combinations of six NWS-derived strains, including two strains of S. thermophilus (St), two strains of L. helveticus (Lh), and two strains of L. delbrueckii subsp. lactis (Ld). Milk fermentations were performed using single monocultures (St, Lh, and Ld), 1:1 co-cultures (St + Lh, St + Ld, and Lh + Ld), and 1:1:1 tri-cultures (St + Lh + Ld), resulting in a total of 17 inoculation conditions. The experimental design of the cross-feeding tests is illustrated in Figure 1. Specifically, individual strains were initially propagated in tubes containing 5 mL of the corresponding isolation medium at 42°C for 48 h according to the culture conditions described in Supplementary Table 1. Pre-cultures were scaled up to 35 mL under the same conditions and incubated until reaching the stationary phase. Cell densities were spectrophotometrically quantified at 600 nm (OD₆₀₀), using species-specific calibration curves to correlate OD values with viable cell counts (CFU/mL) (Rutella et al., 2016). Prior to inoculation, appropriate volumes of each culture were centrifuged (6,000 rpm, 10 min, 4°C), and cell pellets were resuspended in 1 mL of sterile physiological solution. Cell suspensions were then inoculated in 40 mL of partially skimmed UHT milk (Parmalat, Reggio Emilia, Italy), previously aseptically aliquoted in 100 mL screw-cap Erlenmeyer flaks, to achieve a final concentration of app. 2 × 10⁷ CFU/mL. For co-culture and tri-culture fermentations, cell suspensions of the respective species were combined in 1:1 or 1:1:1 ratio immediately prior to inoculation to reach the same final concentration (2 × 10⁷ CFU/mL). The total inoculum size was kept constant across all conditions to avoid variability in acidification rate due to differences in inoculum density. Uninoculated milk was included as a negative control in each experiment. All milk fermentation assays were performed in triplicate. Inoculated milk samples were incubated at 42°C for 160 h in a temperature-controlled water bath.

Milk acidification was monitored three times per day (at 9:00, 13:00, and 17:00) by aseptically collecting 800 μL of sample and measuring pH using a penetration pH meter (XS Instruments, Carpi, MO, Italy). Fermentations were considered complete when the pH remained constant for at least three consecutive measurements. Changes in pH (ΔpH), calculated as the difference between the initial pH (pH₀) and the pH measured at each time point, were plotted against time to generate the acidification curves. Acidification curves were modeled using the gcFitSpline function available in grofit R package (Kahm et al., 2010) to calculate maximum acidification rate (μ_max. pH units h^–1) and maximum acidification efficiency (expressed as ΔpH_max).

At the end of fermentation, fermented milk samples were centrifuged (9,000 rpm, 10 min, 4°C), and the supernatants were 0.45 μm-filtered before enzymatic determination of D- and L-lactic acid concentrations, according to the manufacturer’s instructions (Cat. No.: K-DLATE; Megazyme, Wicklow, Ireland). Enzymatic measurements were performed using a microplate reader (Multiskan SkyHigh, Thermo Fisher Scientific, United States).

Statistical analyses were performed using GraphPad Prism software (v.10, GraphPad Software, La Jolla, CA, United States). Data were analyzed by two-way ANOVA, followed by Tukey’s multiple comparison test as post hoc analysis, unless otherwise specified. Differences were considered statistically significant at p < 0.05 and, where applicable, are indicated by different letters.

The physicochemical and microbiological properties of the two NWS samples analyzed in this study are reported in Table 2. TMC values exceeded 9 Log₁₀ cells/mL in both samples, with viability values of 90.4% for type-D NWS and 89.5% for type-H NWS. Lactobacilli counts on conventional MRS medium were 8.05 Log₁₀ CFU/mL for type-D NWS and 8.01 Log₁₀ CFU/mL for type-H NWS. In both samples, streptococci (evaluated on M17-SSW medium) and contaminant yeasts (evaluated on YPDA medium) were present, in agreements with previous reports (Martini et al., 2021; Sola et al., 2022). Overall, the data indicate that the NWS samples are high dense cultures of viable and metabolically active LAB cells.

Metataxonomic profiles were determined for type-D and type-H NWS samples. Both communities were dominated by L. delbrueckii and L. helveticus, cumulatively accounting for > 70% of the total microbial abundance. The type-D sample was enriched in L. delbrueckii (78.62%), followed by L. helveticus (17.14%), S. thermophilus (3.96%), and L. fermentum (0.28%), whereas the type-H NWS in L. helveticus (79.81%), followed by L. delbrueckii (14.03%), S. thermophilus (6.10%), and L. fermentum (0.06%) (Figure 2). The results were consistent with the previous classification of these NWS cultures in type-D and type-H communities (Sola et al., 2022).

The reduced viability of the NWS lactobacilli may be related to nutritional limitations, such as auxotrophies for several amino acids and growth factors, which are commonly reported in LAB species. To evaluate this hypothesis, 11 different MRS-based media were tested for the ability to support NWS lactobacilli growth, assessed both by plate counts (Log₁₀ CFU/mL) and by the capability to maintain isolates viable ex situ over time as axenic culture. Five nitrogen and growth factors sources were used as supplements: caseins hydrolysate (1, 2, and 4% w/v), tryptone (1, 2, and 4% w/v), lactoalbumin hydrolysate (1, 2, and 4%), YE (4% w/v), and SSW (7% w/v). For each supplementation, type-D NWS and one type-H NWS samples previously characterized by metagenomics were submitted to microbiological analyses.

Concerning plate counts, supplementations with CAS, YE, and TRYPTO were the most effective supplements in supporting the growth of NWS lactobacilli. Specifically, caseins supplementation increased plate counts compared to the control condition (MRS without supplementations; p < 0.05) (Figure 3A). In type-D NWS sample, this effect was concentration-dependent, with caseins 4% yielding the highest Log₁₀ CFU/mL values, whereas in type-H sample, 1% caseins supplementation resulted in the highest plate counts (Figure 3A).

A similar trend was observed for tryptone supplementations. In the type-D sample, plate counts increased proportionally with tryptone concentration compared to the control (p < 0.05), whereas in the type-H sample, 1% tryptone was enough to significantly enhance plate counts relative to MRS medium (Figure 3B).

Lactalbumin hydrolysate also significantly improved plate counts compared to the control, but only at the highest concentrations tested in both type-D and type-H samples (Figure 3C).

By contrast, YE and SSW supplementations produced divergent effects: YE significantly increased plate counts compared to the control, whereas SSW supplementation failed to promote lactobacilli growth in both NWS samples (Figure 3D).

The 7 most effective supplements in supporting plate counts were subsequently selected to evaluate their capability to support cultivability of NWS lactobacilli, assessed as the percentage of viable axenic cultures after three successive purification steps. Remarkably, the supplements that have maximized plate counts were largely ineffective in sustaining the cultivability during axenic propagation (Figure 4). By contrast, MRS supplemented with 4% (v/v) lactalbumin hydrolysate resulted in the highest cultivability for isolates from both type-D and type-H NWS, although it failed to enhance plate counts. For isolates from type-D NWS, supplementation with 4% tryptone yielded cultivability levels comparable to those observed with 4% lactalbumin, whereas YE supplementation maintained cultivability in approximatively 20% of isolates, similar to the control condition. For the type-H NWS isolates, 4% lactalbumin resulted in the highest cultivability, followed by 2% caseins, and YE (Figure 4). Across most culture conditions, the highest mortality rate was observed during the first purification step for both the sets of isolates (data not shown).

To evaluate whether supplementation of MRS medium with different nitrogen sources affected species representativeness, isolates were identified at species level using DNA barcoding methods. As shown in Table 3, the majority of lactobacilli isolated from both type-D and type-H NWS samples were identified as L. helveticus, suggesting that all tested media largely failed to support the cultivation of L. delbrueckii, even when this species dominated the microbial community of type-D NWS, as revealed by metataxonomic analysis (Figure 2). Remarkably, only MRS-SSW and MRS-LACTO 4% enabled the isolation of a limited number of L. delbrueckii strains and maintained them in a viable and cultivable state. In contrast, MRS-YE supported the isolation of L. helveticus and L. fermentum only (Table 3). Nevertheless, none of the tested media reflected the actual species composition observed in type-D NWS (Figure 2). Taken together, the results indicate that supplementation strategies that effectively increased plate counts, such as MRS-YE, MRS-TRIPTO 2 and 4%, MRS-CAS 2 and 4%, were not suitable for cultivating L. delbrueckii. This outcome is likely attributable to the reduced survival of L. delbrueckii during subsequent purification steps, rather than to insufficient initial growth on these media.

As members of a complex microbial community, NWS lactobacilli may be poorly adapted to grow under axenic conditions, where microbial interactions and metabolite exchanges are absent. In Swiss cheese NWS cultures, formic acid and folate are produced by S. thermophilus and used by L. delbrueckii (Somerville et al., 2022). This metabolic interaction is also a well-known feature of the protocooperation between S. thermophilus and L. delbrueckii subsp. bulgaricus in yogurt cultures (Sieuwerts et al., 2008, 2010). Furthermore, cysteine supplementation has been reported to enhance the growth of several oxygen-sensitive lactobacilli (Dave and Shah, 1998; Soto et al., 2019; Serrador et al., 2025). Based on these considerations, the effects of cysteine, formic acid, and folate supplementations on plate counts and cultivability of NWS lactobacilli were investigated. Specifically, C medium, containing L-cysteine (0.05% w/v), CF medium, containing L-cysteine (0.05% w/v) and formic acid (5 mM), and CFFa medium, containing L-cysteine (0.05% w/v), formic acid (5 mM), and folic acid (226 μM), were evaluated for the ability both to support plate counts of NWS lactobacilli and to maintain isolates from both type-D and type-H NWS in a viable and cultivable state.

MRS-YE medium was confirmed to be the most effective for plate count recovery in both NWS samples (p < 0.05) (Figure 5). By contrast, addition of cysteine, formic acid, and folic acid produced distinct effects in the two NWS samples. In type-D NWS, the highest plate counts were scored on CF medium, which exhibited values comparable with those observed on MRS and MRS-YE. Conversely, in type-H NWS, all tested conditions showed significantly lower Log₁₀ CFU/mL values compared with the control MRS medium (p < 0.05) (Figure 5). These differences may reflect the distinct microbial composition of the two NWS cultures, with type-D NWS being characterized by a higher abundance of L. delbrueckii, and type-H NWS dominated by L. helveticus (Figure 2). The results suggest that supplementation with cysteine and formic acid may favor the growth of L. delbrueckii, which was present at high abundance in type-D NWS.

Cultivability results were consistent with the trends observed for plate count recovery. As shown in Figure 6, supplementation with cysteine and formic acid resulted in 100% cultivability of isolates derived from type-D NWS. In contrast, for type-H NWS, cultivability values obtained with C, CF, and CFFa media were significantly lower than those observed with MRS and MRS-YE (p < 0.05). Overall, these findings support that supplementation with cysteine and formic acid represents the most effective condition for both enumeration and isolation of the lactobacilli fraction from type-D NWS enriched in L. delbrueckii.

The ability of C, CF, and CFFa media to isolate the species present within type-D or type-H NWS communities was evaluated by performing species identification of axenic cultures obtained from each culture condition. For isolates derived from type-D NWS, CF medium was the only condition that supported the maintenance of L. delbrueckii isolates in a viable and cultivable state (Table 4). However, L. delbrueckii accounted for only 35% of the isolates recovered on CF medium, indicating that the dominance of L. delbrueckii has not been completely represented under this cultivation condition. Interestingly, all L. delbrueckii isolates belonged to the subspecies lactis, as determined by 16S-ARDRA with EcoRI (Giraffa et al., 1998) and confirmed by 16S rRNA gene sequencing. In contrast, for isolates obtained from type-H NWS, MRS-YE was the only medium that allowed the recovery of both L. helveticus and L. fermentum. None of the tested media allowed for the isolation of the minor L. delbrueckii fraction present in type-H NWS (Table 4).

To further confirm the positive effect of cysteine and formic acid supplementation on the growth of L. delbrueckii subsp. lactis, growth performance of two representative strains (C5I3 and T1104) was evaluated in SSW, C, CF, and CFFa media and expressed as a percentage relative to growth in MRS medium used as the control. While SSW supplementation decreased the growth percentage of both strains, C, CF, and CFFa exerted a positive effect, with C and CF representing the most effective conditions (p < 0.05) (Supplementary Figure 1).

To test whether the presence of other species could specifically promote the growth and activity of L. delbrueckii, cross-feeding experiments were carried out by fermenting milk with different combinations of mono-, co-, and tri-cultures of L. helveticus, L. delbrueckii subsp. lactis, and S. thermophilus strains isolated from type-D and type-H NWS communities. Specifically, L. helveticus (Lh) strains C001.15 and LBB04, L. delbrueckii subsp. lactis (Ld) strains T1104 and C5I3, and S. thermophilus strains (St) RBC06 and C001.27 were combined into three inoculation sets: set 1 (C001.15, RBC06, and T1104), set 2 (LBB04, C001.27, and C5I3), and set 3 (C001.15, RBC06, and C5I3) (Figure 1). Each strain was inoculated into milk as monocultures, in pairwise co-cultures, and tri-cultures. Acidification kinetics (Supplementary Figure 2) were analyzed to estimate the maximum acidification rate (μ_max) and maximum acidification efficiency (ΔpH_max).

In monoculture, Lh exhibited the highest acidification rates, followed by Ld and St, respectively (p < 0.05) (Figure 7A). Both Ld monocultures showed intermediate μ_max values relative to Lh and St monocultures (Figure 7A). Across all inoculation sets, tri-cultures consistently displayed the highest μ_max values (p < 0.05), followed by the co-cultures St + Lh and St + Ld.

Remarkably, in all strain combinations tested, co-inoculation of milk with Lh and Ld resulted in reduced acidification rates compared with the corresponding monocultures, regardless of the specific Ld and Lh strains used (Figure 7A). By contrast, in St + Ld co-cultures, μ_max values were significantly higher than those observed in the corresponding Ld and St monocultures (p < 0.05) (Figure 7A), indicating a mutual benefit from co-cultivation. A similar enhancement was observed in Lh + St co-cultures, where all strain combinations exhibited significantly higher acidification rates than their respective monocultures (p < 0.05) (Figure 7A). Interestingly, in inoculation set 2, co-cultivation of St strain C001.27 with Ld strain C5I3 resulted in μ_max values comparable to those observed under tri-culture conditions. In contrast, co-cultivation with Lh strain LBB04 yielded significantly lower μ_max values compared to the corresponding St + Ld interaction. However, these values remained comparable to those recorded for the co-culture condition in other inoculation sets (p < 0.05) (Figure 7A).

Statistical analysis of the maximum acidification efficiency (ΔpH_max) largely confirmed the trends described for μ_max, with the exception of Ld strain T1104 in inoculation set 1 (Figure 7B). Strain T1104 showed lower ΔpH_max than the conspecific strain C5I3 and exhibited a such limited acidification ability in monoculture that co-cultivation with Lh strain C001.15 resulted in lower final pH values and, consequently, a higher ΔpH_max compared with the T1104 monoculture (p < 0.05). Nonetheless, even under these conditions, Lh + Ld co-cultures remained less effective in acidification than Lh monoculture (p < 0.05).

Overall, the findings support the presence of negative interactions between L. helveticus and L. delbrueckii subsp. lactis and positive interactions between L. delbrueckii subsp. lactis and S. thermophilus, as well as between S. thermophilus and L. helveticus.

Since S. thermophilus mainly produces L-lactate (Ghailan and Niamah, 2025), L. delbrueckii subsp. lactis produces D-lactate (Bernard et al., 1991), and L. helveticus produces a racemic mixture of L- and D-lactate (Kylä-Nikkilä et al., 2000), the D/L-lactate ratio at the end of milk fermentation can provide insights into species interactions during co-cultivation. Accordingly, milk samples collected at the end of each fermentation were analyzed to determine D- and L-lactate concentrations.

In monoculture, Lh strains produced a racemic mixture of lactic acid, whereas Ld strains produced exclusively D-lactate, as expected (Figure 8). However, the limited growth of Ld strain T1104 resulted in a very low level of D-lactate. St strains RBC06 and C001.27 predominately produced L-lactate, which accounted for 67.2 and 83% of total lactate, respectively. Regardless of the lactate enantiomer produced, monocultures produced a very low amount of total lactate, consistent with the slow acidification phenotype previously reported for NWS isolates (Morelli et al., 1986; Fortina et al., 1998; Hebert et al., 2001).

Across all three strain combinations, total lactic acid production was highest in tri-cultures, followed by St + Lh and St + Ld co-cultures (p < 0.05) (Figure 8), consistent with the acidification kinetics. Although absolute lactate concentrations varied among inoculation sets, D-lactate and L-lactate were produced in comparable amounts in all tri-cultures, indicating that Ld strains remained viable and metabolically active. In two of the three sets (sets 1 and 2), D-lactate levels in tri-cultures were significantly higher than those detected in the corresponding Ld monocultures (p < 0.05), supporting the presence of positive interactions that enhanced Ld activity in mixed cultures.

In milk fermented with St and Lh, total lactate production increased relative to monocultures (Figure 8). Under St + Lh conditions, L-lactate predominates, accounting for 86.73% of total lactate in the co-culture of strains RBC06 and C001.5 and 74.08% in the co-culture of strains C001.27 and LBB04.

Similarly, in St + Ld co-cultures, total lactate concentrations also exceeded those of the respective monocultures, with production dominated by D-lactate, suggesting that acidification was largely attributable to Ld activity (Figure 8).

In inoculation sets 2 and 3, Lh + Ld co-cultures yielded significantly lower total lactic acid than tri-cultures and the St + Lh and St + Ld co-cultures (Figure 8). In both cases, D-lactate accounted for the majority of lactate produced (75.95% in set 2 and 56.63% in set 3, respectively), indicating that Ld was metabolically active. Consistent with ΔpH_max values, an exception was observed in inoculation set 1, where the Lh + Ld co-culture, represented by C001.15 and T1104, produced total lactate amount comparable to those of St + Lh and St + Ld co-cultures (p > 0.05) (Figure 8), but significantly lower than the corresponding tri-culture (p < 0.05) (Figure 8). In this Lh + Ld combination, L-lactate represented 67.42% of total lactate, suggesting that Lh was more metabolically active than Ld. This result is likely attributable to the very slow acidification phenotype of Ld strain T1104.