A Study on the Stability and Carbohydrate Metabolic Traits of Starter Cultures in Response to Continuous Subculturing

Abstract

The industrial application of starter cultures requires stable physiological and genetic performance. In this study, Streptococcus salivarius subsp. thermophilus and Lactobacillus delbrueckii subsp. bulgaricus were continuously subcultured. Physiological stability was assessed through colony morphology, fermentation activity, and growth profiling. Genetic stability was evaluated through comparative genomics of carbohydrate metabolism networks and single-nucleotide polymorphism (SNP) analysis. The results showed that after 2000 generations, the cellular morphology of the strains remained intact. Additionally, the strains exhibited enhanced growth performance and fermentation capability. The Gompertz model revealed that adapted S. thermophilus A37 and L. bulgaricus B29 exhibited shortened lag phases, increased maximum specific growth rates, and high stationary-phase cell densities. Phenotypic microarray and comparative genomics revealed that S. thermophilus mainly used mono- and disaccharides, with impaired ribose metabolism due to the absence of the rbsk gene in the pentose phosphate pathway. In contrast, L. bulgaricus metabolized diverse oligosaccharides, sugar alcohols, and plant-derived substrates. Additionally, it effectively catabolized ribose through the phosphoketolase pathway and possessed a trehalose degradation cluster. All strains exhibited genomic stability, with SNPs revealing fewer than 21 variations per isolate. This study provides an important theoretical foundation for evaluating the stability of fermentation starter cultures.

1. Introduction

Yogurt, a traditional fermented dairy product with a long history of consumption, has become a staple in the human diet owing to its distinctive flavor, smooth texture, and health benefits [1,2]. Yogurt is traditionally produced through the symbiotic fermentation of milk by Streptococcus salivarius subsp. thermophilus (S. thermophilus) and Lactobacillus delbrueckii subsp. bulgaricus (L. bulgaricus) [3]. In industrial applications, starter cultures are routinely subjected to prolonged preservation and repeated subculturing. During these processes, strains may accumulate genetic variations, undergo metabolic remodeling, and exhibit phenotypic changes, all of which can influence their fermentation performance. Furthermore, during production, strains encounter various environmental stresses, such as temperature fluctuations, osmotic shock, and pH variations [4]. These stresses act as selective pressures that can enrich for pre-existing genomic variations and induce adaptive regulatory responses, leading to altered metabolic regulation and phenotypic outcomes in lactic acid bacteria, thereby affecting their fermentation performance [5,6,7]. Therefore, developing an effective and stable evaluation system is crucial for the selection and application of fermentation starter strains.

Current methods for evaluating the stability of lactic acid bacteria strains are as follows: (1) cultivation under variable environmental conditions (e.g., temperature, pH, or nutrient composition) [8,9]; (2) exposure to extreme stress conditions, such as acidic, thermal, oxidative, or dehydrated environments [10]; and (3) continuous subculturing. Among these methods, continuous subculturing is widely used for assessing the physiological and genetic stability of lactic acid bacteria [11]. During long-term serial passage, strains accumulate genomic mutations and undergo transcriptional regulatory remodeling, thereby driving adaptive changes in metabolic pathways, carbon utilization efficiency, and environmental fitness [12]. Compared with short-term stress treatments, continuous subculturing under relatively stable growth conditions minimizes the interference of the confounding effects of transient environmental fluctuations [13]. This strategy facilitates a systematic evaluation of genetic architecture and physiological traits over long-term continuous subculturing [14]. However, most existing studies on the stability of lactic acid bacteria (LAB) have mainly focused on short-term stress responses, while the metabolic dynamics and genetic evolution of LAB under long-term serial passage remain unclear.

In this study, we systematically assessed the physiological and genetic stability of starter cultures under long-term serial passage conditions. Five S. thermophilus strains and three L. bulgaricus strains were continuously subcultured for 2000 generations. Morphological changes were first examined using scanning electron microscopy (SEM). In parallel, fermentation activity, growth kinetics, and metabolic capacity were assessed to comprehensively investigate the effects of prolonged cultivation on strain physiology. Subsequently, whole-genome sequencing was performed, followed by comparative genomic analyses to analyze characteristic carbon source metabolic networks and SNP variations, thereby revealing potential genetic changes. Overall, this study aimed to elucidate the stability patterns of S. thermophilus and L. bulgaricus under long-term passage conditions, providing theoretical insights to support their industrial application as fermentation starter cultures.

2. Results and Discussion

2.1. Morphological Changes in Fermentation Strains During Continuous Subculture

The colony and cell morphology of S. thermophilus and L. bulgaricus were monitored throughout serial subculturing. Colonies of wild-type S. thermophilus displayed a milky-white to grayish-white appearance, with diameters ranging from 0.5 to 2 mm. They appeared round, convex, and well-defined and exhibited smooth and moist surfaces (Figure 1a,g and Figure S1). The diameter of colonies of wild-type L. bulgaricus was 1–3 mm, appearing as round, convex, milky-white colonies with smooth and moist surfaces (Figure 1m,s and Figure S1). Regarding cellular morphology, S. thermophilus mainly formed short- to medium-length chains (Figure 1c,i and Figure S1), whereas L. bulgaricus predominantly occurred in pairs, clusters, or short chains (Figure 1o,u and Figure S1). After 2000 generations of continuous subculturing, both the colony and cellular morphology of the strains remained unchanged. S. thermophilus colonies retained their milky-white to grayish-white color, diameter (0.5–2 mm), and smooth, convex, well-defined structure (Figure 1b,h and Figure S1), while the cells continued to form short- to medium-length chains (Figure 1d,j and Figure S1), showing no significant differences compared with the wild-type strains. Similarly, L. bulgaricus colonies remained round, convex, milky-white, and 1–3 mm in diameter, with smooth and moist surfaces (Figure 1n,t and Figure S1). Cells still occurred predominantly in pairs, clusters, or short chains (Figure 1p,v and Figure S1), consistent with the wild-type strains.

Morphological alterations during prolonged serial subculturing were further examined using SEM. After 2000 generations, the wild-type S. thermophilus A1 shifted from short chains composed of 2–5 cellular units to medium-long chains comprising 10–15 units (Figure 1e,f). In contrast, S. thermophilus A4, A31, A37, and A72 remained stable, retaining their original chain-length architecture (Figure 1k,l and Figure S1). Previous reports have shown that chain length in streptococci may be regulated by extracellular polysaccharides or autolysins. For example, Shibata et al. [15] identified the atlA gene in Streptococcus mutans, which encoded the major autolysin AtlA. The disruption of AtlA significantly reduced autolytic activity and resulted in impaired daughter cell separation, thus elongating cellular chains. SEM results for L. bulgaricus revealed that cells generally displayed smooth, well-defined surfaces and occurred primarily as single cells or short chains. However, a morphological change was observed in the specific wild-type strain L. bulgaricus B29. Initially, this strain exhibited visible cell wall defects and fractures (Figure 1q). Remarkably, after 2000 generations of serial subculture, its cellular morphology appeared more intact, with previously fractured surfaces showing signs of restoration, resulting in smoother and more defined structures (Figure 1r). This phenomenon indicates that prolonged subculturing under stable conditions contributed to enhanced cell wall integrity in this particular strain. In contrast, no such morphological deterioration or recovery was observed in the wild-type strains L. bulgaricus B39 and B43; these strains retained intact cellular morphology throughout 2000 passages (Figure 1w,x and Figure S1). The distinct morphological trajectory observed in strain B29 may be attributed to adaptive modifications in cell wall constituents, such as peptidoglycan, teichoic acids, and surface-associated proteins, which are known to modulate structural integrity [16]. This adaptive response is consistent with findings in other lactic acid bacteria, such as Lactiplantibacillus plantarum, which can upregulate the biosynthesis of structural cell wall components to mitigate cellular damage under stress [17].

2.2. Assessment of Growth Characteristics and Fermentation Activity After Continuous Subculture

After S. thermophilus and L. bulgaricus were continuously subcultured, the growth characteristics and fermentation performance of the strains were evaluated by measuring the pH of the culture medium and OD600 values. In wild-type S. thermophilus strains (A1, A4, A31, A37, and A72), the pH values ranged from 4.71 to 4.81, with OD600 values between 1.39 and 1.50 (

Table 1. Growth characteristics of S. thermophilus and L. bulgaricus in the wild-type strain and generation 2000 during continuous subculturing. One viability unit (U) is defined as 1 μmol of lactic acid per 1 × 10⁷ CFU of bacterial cells during the fermentation of 1 mL of milk under the above conditions. All assays were performed in triplicate, and data are expressed as mean ± standard deviation (SD). Values in the same row that share a lowercase letter (a, b) are not significantly different; different letters indicate statistically significant differences (p < 0.05) between 0 generations and after 2000 generations of the same strain, as determined by one-way ANOVA followed by Tukey’s multiple comparison test.

Generation Time (n)	pH Values		OD600		Fermentation Activity (U)
	0	2000	0	2000	0	2000
A1	4.73 ± 0.03 a	4.69 ± 0.01 a	1.37 ± 0.02 a	1.49 ± 0.01 b	45.33 ± 1.02 a	51.37 ± 0.80 b
A4	4.81 ± 0.01 a	4.74 ± 0.01 a	1.49 ± 0.01 a	1.52 ± 0.01 b	57.41 ± 0.98 a	64.10 ± 2.06 b
A31	4.75 ± 0.00 a	4.66 ± 0.02 b	1.39 ± 0.01 a	1.39 ± 0.02 a	50.27 ± 1.56 a	56.29 ± 1.14 b
A37	4.81 ± 0.00 a	4.64 ± 0.01 b	1.48 ± 0.02 a	1.53 ± 0.01 b	50.86 ± 0.91 a	58.95 ± 1.64 b
A72	4.71 ± 0.00 a	4.66 ± 0.01 b	1.50 ± 0.00 a	1.55 ± 0.02 a	52.10 ± 1.28 a	55.65 ± 0.51 b
B29	4.30 ± 0.00 a	4.07 ± 0.01 b	2.10 ± 0.01 a	2.19 ± 0.01 b	57.01 ± 0.42 a	62.41 ± 0.67 b
B39	4.32 ± 0.01 a	4.31 ± 0.01 a	2.00 ± 0.01 a	2.06 ± 0.01 b	33.10 ± 0.78 a	41.69 ± 1.55 b
B43	4.48 ± 0.01 a	4.33 ± 0.01 b	1.99 ± 0.01 a	2.08 ± 0.01 b	56.38 ± 2.01 a	61.90 ± 1.72 b

). After 2000 generations of subculturing, the pH value ranged from 4.64 to 4.74, and OD600 values were between 1.39 and 1.55 (Table 1). Notably, the pH of S. thermophilus A37 significantly decreased from 4.81 to 4.64, while its OD600 significantly increased from 1.48 to 1.53 (p < 0.05) compared with the wild-type strain.

In contrast, the wild-type L. bulgaricus strains (B29, B39, and B43) showed lower medium pH values, ranging from 4.30 to 4.48, and higher OD600 readings between 1.99 and 2.10 (Table 1). After 2000 generations of subculturing, these strains demonstrated a further decrease in medium pH (4.07–4.33) and an increase in OD600 (2.06–2.19), indicating enhanced growth and acidification capacity relative to the S. thermophilus group. Among them, L. bulgaricus B29 exhibited a pH decrease of 0.23 and an increase in OD600 (0.09). Although the absolute pH difference may appear numerically small, a shift of 0.23 units corresponds to a substantial change in proton concentration, reflecting a physiologically meaningful enhancement in acid production. Collectively, the results indicate that long-term subculturing improved both acidification capacity and biomass accumulation in the evaluated strains.

Fermentation activity significantly increased in both species after 2000 generations (Table 1) (p < 0.05). Notably, after 2000 passages, the fermentation activity of S. thermophilus (A4 and A36) exceeded 60 U, while that of L. bulgaricus B29 and B43 was approximately 62.41 U and 61.9 U, respectively, indicating excellent fermentation performance. Prolonged adaptation of wild-type starter cultures in lactose-containing media can change key enzyme activities related to metabolic regulation and energy utilization efficiency, thereby enhancing environmental adaptability. For example, Yu et al. reported that elevated β-galactosidase and urease activities enhanced lactose metabolism in different S. thermophilus strains, improving fermentation performance [18].

2.3. Carbohydrate Metabolic Profiling

Phenotypic MicroArray (PM) technology, a high-throughput platform that enables systematic assessment of cellular phenotypes such as carbon source utilization and metabolic activity, was used on the OmniLog™ system to comparatively analyze the carbon utilization profiles of L. bulgaricus across 46 distinct carbon sources. By analyzing the carbon utilization profiles of 2000 generations of strains, it was found that S. thermophilus utilized 28 carbon sources, while L. bulgaricus metabolized 43 carbon sources (Figure 2). The utilization profiles were clearly segregated into three distinct metabolic clusters (Clusters I–III).

Cluster I, co-driven by S. thermophilus and L. bulgaricus, consisted of monosaccharides, disaccharides, and metabolic intermediates. Monosaccharides in Cluster I included fundamental hexoses, such as D-glucose, D-galactose, and D-mannose, and pentoses, such as D-xylose, D-ribose, L-arabinose, and L-lyxose (Figure 2). In addition, Cluster I contained monosaccharide epimers, including epimers of glucose or fructose (e.g., D-allose, D-psicose, and D-tagatose) and 6-deoxy rare hexoses (e.g., L-fucose, D-fucose, and L-rhamnose) (Figure 2). Notably, S. thermophilus effectively metabolized L-arabinose, L-rhamnose, L-lyxose, and D-xylose, whereas L. bulgaricus weakly utilized these sugars (Figure 2). This difference might be attributed to the distinct ecological adaptations and genomic metabolic potentials of the two species. Cluster I contained some metabolic intermediates, such as D-fructose-6-phosphate and dihydroxyacetone. Dihydroxyacetone can be phosphorylated by dihydroxyacetone kinase and subsequently enter the EMP pathway for further metabolism [19].

Cluster II is mainly driven by L. bulgaricus and comprises oligosaccharides (e.g., dextrin, stachyose, maltotriose), functional disaccharides (e.g., maltose, turanose, lactulose), and sugar alcohols (Figure 2). The carbon sources in Cluster II are predominantly derived from plant or starch sources, and their utilization typically depends on starch hydrolysis systems involving key enzymes, including α-glucosidase [20,21] and trehalose-6-phosphate hydrolase [22]. Notably, D-trehalose serves as a carbon source and significantly enhances bacterial tolerance to processing stress, dehydration, and freezing conditions [23,24]. The carbon source utilization profile of Cluster II highlights the significant advantage of L. bulgaricus strains in metabolizing sugar alcohols and oligosaccharides, reflecting their adaptive capacity to complex environmental substrates.

Cluster III, distinct from clusters I and II, represents strains with specialized substrate-utilization capabilities, such as L. bulgaricus B43. This strain can utilize D-melibiose, L-galactonic acid-δ-lactone, D-galacturonic acid, and phenylethylamine (Figure 2). L-galactonic acid-δ-lactone is a characteristic intermediate produced during pectin decomposition, while D-galacturonic acid is a primary component of plant cell wall pectin. In prokaryotes, D-galacturonic acid is catabolized through two principal pathways: an oxidative route and an isomerization pathway [25,26]. Additionally, we observed that the carbohydrate utilization profiles of S. thermophilus and L. bulgaricus remained stable between wild-type strains and 2000 generations, indicating that their carbohydrate metabolic capabilities remained consistent.

2.4. Analysis of Microbial Growth Kinetics

Based on morphological observations, growth performance, fermentation activity, and carbohydrate metabolism profiling, S. thermophilus A37 and L. bulgaricus B29 were selected for microbial growth kinetics analysis. Both strains exhibited adaptive changes after long-term serial passage. The OD600 values of both wild-type (generation 0) and 2000-generation passaged strains were monitored at 1 h intervals to assess their growth curves. The Gompertz model was used to fit the growth curves of wild-type and 2000-generation passaged strains of S. thermophilus A37 and L. bulgaricus B29. The adjusted coefficient of determination (adjusted R²) and the root mean square error (RMSE) were used to assess the model performance. As shown in

Table 2. Growth kinetic parameters of S. thermophilus A37 and L. bulgaricus B29 were determined in both wild-type strains and after 2000 generations of continuous subculturing. Growth dynamics were modeled using the Gompertz equation. The analyzed parameters included N_max (maximum cell density, corresponding to OD600), R_m (maximum specific growth rate), λ (lag phase duration, h), N₀ (initial cell density), root mean square error (RMSE), and adjusted R² (adjusted coefficient of determination). Bacterial counts were measured at different cultivation times (0, 2, 3, 5, and 8 h). All experiments measuring viable cell counts were performed in triplicate, and data are expressed as mean ± standard deviation (SD). Within the same row, different lowercase letters (a, b) indicate significant differences in viable cell counts at different time points for the same strain (p < 0.05).

Gompertz	A37-0	A37-2000	B29-0	B29-2000
Nmax	1.089	1.203	1.837	1.998
Rm	0.536	0.691	0.634	0.887
λ (h)	5.323	4.688	5.043	4.427
N0	0.026	0.050	0.061	0.087
SSE	0.005	0.007	0.013	0.015
Adjusted R2	0.996	0.996	0.997	0.997
RMSE	0.025	0.031	0.043	0.047
0 h (log10 CFU/mL)	6.77 ± 0.01 a	6.80 ± 0.02 a	5.43 ± 0.01 a	5.51 ± 0.02 a
2 h (log10 CFU/mL)	6.71 ± 0.05 a	6.89 ± 0.02 b	5.75 ± 0.03 a	6.15 ± 0.01 b
3 h (log10 CFU/mL)	7.20 ± 0.05 a	7.32 ± 0.02 b	6.24 ± 0.04 a	6.78 ± 0.02 b
5 h (log10 CFU/mL)	7.90 ± 0.05 a	8.38 ± 0.05 b	7.15 ± 0.06 a	7.64 ± 0.04 b
8 h (log10 CFU/mL)	8.50 ± 0.02 a	8.70 ± 0.07 b	7.75 ± 0.05 a	7.89 ± 0.03 b

, all strains exhibited a high goodness of fit, with adjusted R² values exceeding 0.995 (ranging from 0.996 to 0.997) and consistently low RMSE values ranging from 0.025 to 0.047. These results show that the Gompertz model accurately captures the growth dynamics of both species under the tested conditions and can be reliably used to compare kinetic parameters between wild-type and long-term subcultured strains. Growth curves of S. thermophilus A37 and L. bulgaricus B29, both in their wild-type and after 2000 generations of continuous subculturing, were obtained during monoculture in MRS medium with lactose substituted for glucose. The growth curves of both strains exhibited a characteristic asymmetric sigmoidal shape, with clearly distinguishable lag, exponential, and stationary phases (Figure 3 and Figure S1). For S. thermophilus A37, the lag phase duration (λ) decreased from 5.323 h in the wild-type strain to 4.688 h in the passaged strain. The corresponding viable counts at 3 h were 7.32 log₁₀ CFU/mL in the passaged strain, compared with 7.20 log₁₀ CFU/mL in the wild-type strain (Table 2). After entering the exponential phase, the maximum specific growth rate (R_m) of the subcultured strain increased from 0.536 to 0.691, indicating accelerated growth (Table 2). Viable counts of the passaged strain were 8.38 log₁₀ CFU/mL, significantly higher than that of the wild-type strain (7.90 log₁₀ CFU/mL; p < 0.05) (Table 2). In the stationary phase, the maximum cell density (N_max) was 1.203, with peak viable counts observed at 8 h (8.70 log₁₀ CFU/mL for the passaged strain vs. 8.50 log₁₀ CFU/mL for the wild-type).

The growth of L. bulgaricus B29 was more pronounced after long-term subculture. Compared with the wild-type strain, the lag phase duration (λ) of the 2000 generation strain reduced from 5.043 h to 4.427 h, and the R_m during the exponential phase increased from 0.634 to 0.887, representing a 28.8% increase. This phenomenon was further confirmed by microbial population dynamics. Although initial inoculum levels were similar, at the end of the lag phase (3 h), the viable counts of the passaged strain (6.78 log₁₀ CFU/mL) exceeded those of the wild-type strain (6.24 log₁₀ CFU/mL). During the exponential phase (approximately 5 h), viable counts of the passaged strain were 7.64 log₁₀ CFU/mL, which were significantly higher than the 7.15 log₁₀ CFU/mL recorded for the wild-type strain (p < 0.05). After the stationary phase (8 h), the viable counts of the passaged strain became 7.89 log₁₀ CFU/mL. These results show that long-term continuous subculturing significantly enhances the growth performance of both S. thermophilus A37 and L. bulgaricus B29.

2.5. S. thermophilus and L. bulgaricus Genome Sequences

Genome sequencing of S. thermophilus and L. bulgaricus strains after 2000 generations of continuous subculturing revealed that the genome sizes of S. thermophilus isolates ranged from 1,769,967 to 1,789,462 bp, with GC contents between 38.93% and 38.95% (Figure 4A). In comparison, the genomes of L. bulgaricus strains ranged from 1,776,843 to 1,818,337 bp, with GC contents ranging from 49.81% to 49.85% (Figure 4B). These genomic characteristics are consistent with those reported for other commonly used starter strains. The genomic analysis revealed that S. thermophilus strains contained between 1882 and 1900 predicted coding sequences (CDSs) and 31–43 structural RNAs, including rRNA and tRNA (

Table 3. The general genomic characteristics of five S. thermophilus and three L. bulgaricus strains. The table includes assembly statistics and functional annotations based on COG and KEGG subsystems. Functional comparison of metabolic genes across these strains was performed using Diamond, with an E-value cutoff set at ≤1 × 10⁻⁵.

	A1	A4	A31	A37	A72	B29	B39	B43
General genome feature
Size	1,784,905	1,785,671	1,789,462	1,769,967	1,781,421	1,818,337	1,812,629	1,776,843
GC content	38.95	38.95	38.94	38.93	38.99	49.84	49.85	49.81
CDS No.	1889	1890	1890	1882	1900	1877	1869	1843
Number of RNAs	41	31	37	43	43	90	86	81
Subsystem features
COG gene No.	1574	1572	1567	1552	1595	1486	1485	1489
Percent of All Genes (%)	83.32	83.17	82.91	82.47	83.95	79.17	79.45	80.79
KEGG Pathway Enrichment
Carbohydrate metabolism	109	111	110	106	105	139	123	127
Amino acid metabolism	124	128	124	122	118	142	102	95
Metabolism of other amino acids	25	26	25	26	25	35	31	32
Energy metabolism	46	46	49	46	47	53	65	62
Lipid metabolism	32	33	32	31	30	37	44	44
Nucleotide metabolism	67	67	67	66	68	75	78	79

). Compared with S. thermophilus, L. bulgaricus strains possessed 1843–1877 CDSs and 81–90 structural RNAs (Table 3). No active prophages were detected within the genomic sequences of either S. thermophilus or L. bulgaricus using PHAST. GC content variation is widely believed to be a mutational bias. The genomic GC content of S. thermophilus was significantly lower than that of L. bulgaricus. The genome sequencing results of S. thermophilus were consistent with previous studies. For example, Kapse et al. reported that S. thermophilus MCC0200 possesses a circular chromosome of 1,855,815 bp with an average GC content of 39.1% [27]. Similarly, Alexandraki et al. analyzed 23 fully sequenced S. thermophilus genomes, which ranged from 1.73 to 2.10 Mb with an average GC content of 39.0% [28]. In contrast, L. bulgaricus strains generally have larger and more GC-rich genomes. Shehata et al. reported genome sizes of 1,752,493 bp and 1,759,908 bp with GC contents of 49.80% and 49.90% for strains L. bulgaricus CBC-LB69 and L. bulgaricus CBC-LB8 [29], respectively. Hao et al. found that L. bulgaricus 2038 has a genome length of approximately 1,872,907 bp and an average GC content of 49.68% [30]. Overall, the genomes of L. bulgaricus strains are larger and have higher GC content than those of S. thermophilus.

2.6. Differences in Carbon Source Metabolism Between S. thermophilus and L. bulgaricus

After continuous subculturing of S. thermophilus and L. bulgaricus strains, a comparative genomic analysis was conducted on five S. thermophilus strains (A1, A4, A31, A37, and A72) and three L. bulgaricus strains (B29, B39, and B43), using S. thermophilus ND03 and L. bulgaricus ND02 as reference genomes. The analysis mainly focused on the utilization of carbon sources, with emphasis on the metabolism of D-ribose, L-arabinose, L-rhamnose, D-xylose, and trehalose. S. thermophilus generally retains the genes associated with pentose metabolism and central carbon metabolism, including the pentose phosphate pathway (PPP), such as tktA/B (transketolase), pfk (phosphofructokinase), and fba (aldolase), which encode the corresponding enzymes and collectively constitute the core metabolic module of the PPP (Figure 5A). The catabolic pathways of D-ribose, L-arabinose, and D-xylose converged into the PPP through the intermediate xylulose-5-phosphate (Xu5P). Xu5P was subsequently funneled into central carbon metabolism, where transketolase- and transaldolase-catalyzed reactions converted PPP intermediates into D-fructose-6-phosphate (F6P) through the catalysis of transketolase (tktA/B). F6P was phosphorylated by phosphofructokinase (pfk) to generate D-fructose-1,6-bisphosphate (F1,6BP). Finally, aldolase (fba) catalyzed the cleavage of F1,6BP into glyceraldehyde-3-phosphate and dihydroxyacetone phosphate (Figure 5B). In S. thermophilus, the D-ribose metabolic pathway was facilitated by rpiA and rpe genes. These genes encoded enzymes that catalyzed the conversion of D-ribose-5-phosphate and D-ribulose-5-phosphate into D-xylulose-5-phosphate, which subsequently entered the PPP. However, the rbsK gene was absent in S. thermophilus, thereby limiting its ability to initiate complete ribose catabolism.

In contrast, L. bulgaricus retained a complete D-ribose catabolic gene cluster (rbsK, rpiA, and rpe) (Figure 5A). The resulting metabolic flux sequentially proceeded through phosphorylation, isomerization, and epimerization to generate Xu5P, which was then cleaved by xylulose-5-phosphate phosphoketolase (xfp) into G3P and acetyl-phosphate, both of which entered central pyruvate metabolism (Figure 5B). Additionally, L. bulgaricus retained multiple genes associated with the metabolism of plant-derived sugars, particularly those involved in trehalose utilization, including crr, malH/treC, pgi, pfk, and fba (Figure 5A). Trehalose was imported and phosphorylated by the trehalose-specific PTS permease (crr). The resulting trehalose-6-phosphate was hydrolyzed by trehalose-6-phosphate hydrolase (malH/treC) to glucose and G6P. Then, G6P was sequentially metabolized by the pgi, pfk, and fba gene products, leading to the generation of C3 intermediates that feed into the pyruvate pathway (Figure 5B). Comparative genomic analysis between S. thermophilus and L. bulgaricus revealed distinct pentose metabolism strategies. Although S. thermophilus mainly uses the classic transketolase-dependent PPP, L. bulgaricus uses a more efficient phosphoketolase bypass that directly channels pentoses into central metabolism with fewer enzymatic steps [31], reflecting its evolutionary adaptation to dairy environments. Additionally, L. bulgaricus harbors a trehalose degradation gene cluster, enabling it to utilize the stress-protective sugar to enhance tolerance under adverse conditions, such as acid, cold, or desiccation [32,33].

2.7. Genetic Stability After Continuous Subculture

Whole-genome resequencing (average depth ≈ 1000×) was performed on S. thermophilus strains (A1, A4, A31, A37, and A72) and L. bulgaricus strains (B29, B39, and B43) after 2000 generations of continuous subculturing to identify SNPs and other genetic variations. As shown in

Table 4. Genomic mutation profiles of S. thermophilus and L. bulgaricus in the wild-type strain and generation 2000 during continuous subculturing.

Gene Mutation Type	Strains
	A1	A4	A31	A37	A72	B29	B39	B43
Single-nucleotide polymorphism (SNP)	6	10	5	6	4	10	15	12
Multiple nucleotide polymorphism (MNP)	0	0	2	3	0	1	3	0
Insertion mutation (INS)	2	1	4	3	0	0	0	2
Deletion mutation (DEL)	0	1	0	2	2	1	1	4
Inversion mutation (INV)	0	0	0	0	0	0	0	0
Duplicate mutation (DUP)	0	0	0	0	0	0	0	0
Balanced chromosomal translocation (BED)	0	0	0	0	0	0	0	0
Intergenic genomic variation	0	0	0	0	0	0	0	0

, genomic alterations, including point mutations, insertions, deletions, and multiple nucleotide polymorphisms (MNPs), were detected in all S. thermophilus strains. For example, S. thermophilus A37 exhibited six SNPs, three MNPs, three insertions, and two deletions, whereas S. thermophilus A72 exhibited fewer variations, comprising four SNPs and two deletions. Similarly, L. bulgaricus strains exhibited comparable genomic variations. Among the strains, L. bulgaricus B39 presented the highest number of variations, a total of 15 SNPs, three MNPs, and one deletion, while L. bulgaricus B43 contained 12 SNPs, two insertions, and four deletions. Overall, the total number of genetic variations in S. thermophilus (A1, A4, A31, A37, and A72) and L. bulgaricus (B29, B39, and B43) remained limited after 2000 generations of continuous subculturing. Pightling et al. [34] proposed that fewer than 21 SNP differences in whole-genome sequencing data could be used as a threshold to support the conclusion that two or more bacterial genomes were closely related or identical. Our results meet this criterion, indicating that the subcultured strains retain high genomic identity with their progenitors, thus exhibiting excellent genetic stability under the experimental conditions.

Functional annotation and pathway enrichment analyses indicated that the identified variations (totaling fewer than 20 SNPs across all strains) were distributed among genes involved in diverse cellular processes, including basic metabolism and stress response. Importantly, no mutations were found in key genes or known regulatory elements directly associated with the pentose phosphate pathway (PPP) or the phosphoketolase pathway (PKP). Therefore, the current mutational profile does not support the notion that genetic mutations directly drove adaptive changes in the activity or flux of these specific pathways during long-term subculturing.

3. Materials and Methods

3.1. Continuous Subculturing of S. thermophilus and L. bulgaricus

The strains were inoculated at 2% (v/v) into the MRS medium in which lactose replaced glucose at a pH of 6.5. Serial subculturing was performed, with transfers conducted during the exponential growth phase, until the cumulative number of cell divisions reached approximately 2000 generations. The generation time (g) of Streptococcus salivarius subsp. thermophilus (S. thermophilus) and Lactobacillus delbrueckii subsp. bulgaricus (L. bulgaricus) during the exponential growth phase was calculated according to the following formula:

$$g = {\displaystyle \frac{t·\lg 2}{\lg N_t-\lg N_0}}$$

(1)

where t represents the cultivation time (h), N₀ is the initial viable cell count (CFU/mL), N_t is the viable cell count (CFU/mL) at time t, and g denotes the generation time (h).

3.2. Measurement of pH, TA, and Viable Counts

Optical density (OD₆₀₀) of the bacterial cultures was measured at 600 nm every 400 generations using a UV-2800A spectrophotometer (Unico Instrument Co., Ltd., Beijing, China). The pH of the culture sample was directly measured with a calibrated pH meter (Model S210, Mettler Toledo, Zurich, Switzerland). Viable counts of S. thermophilus and L. bulgaricus were determined by plating appropriate dilutions on M17 agar and MRS agar at an adjusted pH of 5.4, respectively. Plates were incubated at 42 °C for 48 h, and results were expressed as colony-forming units per gram of sample (CFU/g) [35]. All analyses were performed in triplicate using independently cultured samples.

3.3. Fermentation Vitality

Fermentation activity of strains was assessed every 400 passages during continuous subculturing, with all assays performed in triplicate. Activity was assessed by measuring the amount of acid produced during carbohydrate fermentation. Bacterial cells were inoculated into 10 mL of sterile skim milk at a concentration of 1 × 10⁷ CFU/mL and incubated at 42 °C for 24 h. After the incubation was complete, the samples were immediately cooled in an ice bath to terminate fermentation and then diluted with 20 mL of distilled water. The pH of the diluted samples was adjusted to 8.50–8.55 with 0.1 mol/L NaOH, and the endpoint was held for 30 s. The volume of NaOH consumed was recorded to calculate the lactic acid concentration, which was converted into activity units. One activity unit (U) was defined as the amount of lactic acid (1 μmol) produced by 10⁷ cells during the fermentation of 1 mL of skim milk under the described conditions.

3.4. Scanning Electron Microscopy (SEM) Analysis

Subcultured strains were continuously harvested through centrifugation at 1000× g for 10 min. After the subcultured strains were washed twice with 10 mmol/L of phosphate-buffered saline (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, and 2 mM KH₂PO₄; pH 7.4), the cell pellet was resuspended to a final concentration of 1 × 10⁵ CFU/mL. The cell suspension was fixed with 2.5% (v/v) glutaraldehyde and incubated at 4 °C for 24 h. After the mixture was centrifuged, the bacterial pellet was sequentially dehydrated with 30%, 50%, 70%, and 100% ethanol. The samples were lyophilized and subsequently gold-coated under vacuum (4.0 Pa) using an ion sputter coater (Hitachi MC1000, Hitachi High-Tech Corporation, Tokyo, Japan). A cold field-emission scanning electron microscope (Hitachi SU8010) operated at an accelerating voltage of 5 kV and a beam current of 10 μA was used to examine the morphology of the samples.

3.5. Growth Kinetics Modeling

The growth kinetics of S. thermophilus A37 and L. bulgaricus B29 were described using the Gompertz model. The four-parameter model is suitable for fitting asymmetric sigmoidal growth curves, characterized by initial lag, exponential growth, and stationary phases. The modified Gompertz equation was applied as follows:

$$N\left(t\right) = N_0 + N_{\mathrm{m}}*\exp (-\exp ({\displaystyle \frac{R_{\mathrm{m}}*e}{N_{\mathrm{m}}}} + 1\left)\right)$$

(2)

where N(t) is the cell density (OD₆₀₀) at time t (h); N₀ is the initial cell density; N_m is the maximum cell density; λ is the lag phase duration (h); R_m is the maximum specific growth rate (h⁻¹); and e is the base of the natural logarithm.

3.6. Whole-Genome Sequencing and Comparative Genome Analysis

Genomic DNA was extracted using the NEXTFLEX Rapid DNA-Seq Kit according to the manufacturer’s instructions. Drafts and complete genomes of five S. thermophilus and three L. bulgaricus strains were sequenced using the PacBio Sequel (Pacific Biosciences of California, Inc., Menlo Park, CA, USA)and the Illumina NovaSeq PE150 platform (Illumina, Inc., San Diego, CA, USA). Coding sequences (CDSs) were functionally annotated against the non-redundant, Swiss-Prot, Pfam, Gene Ontology, Clusters of Orthologous Genes (COG), and Kyoto Encyclopedia of Genes and Genomes (KEGG) databases with sequence alignment tools using BLASTP (NCBI BLAST+ version 2.17.0), Diamond, and HMMER. For each predicted protein, the annotation of the highest-scoring hit (e-value < 10⁻⁵) was retained. Genome-wide similarity was assessed using OrthoANIu (OrthoANI using USEARCH). Comparative genomic visualization was performed using the BLAST-based Circular Image Generator (v0.95).

3.7. Phenotypic Analyses

Carbon source utilization profiles of five S. thermophilus and three L. bulgaricus strains were analyzed using the Phenotype MicroArray system (Biolog, Hayward, CA, USA). PM1 and PM2A microplates were used to evaluate the metabolic capabilities of the strains toward different carbon sources [36]. Following the manufacturer’s protocol, the growth on MRS agar was suspended in IF-0a inoculating fluid to obtain a cell suspension with 81% light transmittance. Aliquots of 100 μL were dispensed into the wells of 96-well Phenotype MicroArray (PM) plates. Then, the plates were incubated in an OmniLog reader at 42 °C for 72 h. During incubation, the absorbance of the redox-sensitive dye G was automatically recorded every 15 min. Data were processed and analyzed using the OmniLog PM software 2.3.01 (Biolog) according to the manufacturer’s instructions.

3.8. Statistical Analysis and Genome Submission

Functional classification of protein-coding genes was performed using BLASTP (E-value ≤ 10⁻⁵) against NCBI COGs (Clusters of Orthologs), with a minimum identity of ≥30% and coverage of ≥30%. Sequence similarity was detected using BLAST, and multiple sequence alignments were conducted using Clustal 2.1. Putative carbohydrate metabolic pathways were predicted using KEGG. Data from growth characteristics, fermentation vitality, and phenotypic microarray analyses were analyzed using one-way analysis of variance (ANOVA). Each experiment was performed in triplicate, and pairwise comparisons of treatment means were conducted using Tukey’s test at p < 0.05 with Statistica, SPSS (27.0), and GraphPad Prism 8.0. All assembled genomes were submitted to the NCBI GenBank database (https://www.ncbi.nlm.nih.gov/) and assigned accession numbers. Strain designations and GenBank accession numbers were as follows: S. thermophilus A1 (PX945674); S. thermophilus A4 (PX945677); S. thermophilus A31 (PX945682); S. thermophilus A37 (PX945711); S. thermophilus A72 (PX945712); L. bulgaricus B29 (PX944889); L. bulgaricus B39 (PX944890); L. bulgaricus B43 (PX945891). Genome data of the reference genomes used for genome comparison were S. thermophilus ND03 (CP002340) and L. bulgaricus ND02 (CP002341.1).

4. Conclusions

In this study, after continuous subculturing for 2000 generations, both S. thermophilus and L. bulgaricus strains maintained stable colony and cellular morphology and retained high fermentative activity. Carbon utilization profiling revealed that S. thermophilus metabolized 28 carbon sources, including monosaccharides and disaccharides (e.g., L-arabinose), while L. bulgaricus utilized 43 substrates, including oligosaccharides, functional disaccharides, and sugar alcohols. Long-term subculturing significantly enhanced growth kinetics, which were confirmed by shortened lag phases, elevated maximum specific growth rates, and increased stationary-phase cell densities. Comparative genomic analysis revealed divergent metabolic strategies: L. bulgaricus used an efficient phosphoketolase-dependent bypass for pentose catabolism and possessed a trehalose degradation cluster, whereas S. thermophilus relied on the canonical transketolase-driven PPP. Whole-genome resequencing further confirmed that both species exhibited high genomic identity with their ancestral strains after prolonged subculturing, indicating robust genetic stability suitable for industrial applications. However, the mere presence of genes does not necessarily guarantee functional expression or metabolic flux. Further validation through transcriptomic, proteomic, enzymatic, or metabolic flux analyses would provide deeper insight into the active mechanisms underlying these metabolic pathways.