A snapshot study of the microbial community dynamics in naturally fermented cow’s milk

Abstract Natural fermentation of milk is a prerequisite in the production of traditional dairy products and is considered a bioresource of fermentative microorganisms and probiotics. To understand the microbial dynamics during distinct fermentative phases, the roles of different microbes, and the relationship between bacteria and fungi, microbial community dynamics was investigated by culture‐dependent and culture‐independent approaches. Natural, static fermentation of milk induces the formation of the underlying curds and the superficial sour cream (Zuohe in the Mongolian language). From an overall perspective, viable LAB increased remarkably. Yeast showed an initial increase in their abundance (from 0 hr to 24 hr), which was followed by a decrease, and mold was detected at the later stages of fermentation (after 68 hr). The observed trends in microbiota variation suggest an antagonistic interaction between bacteria (LAB) and fungi (yeast and mold). The beneficial bacterial and fungal genus and species (e.g., Lactococcus, Streptococcus, Leuconostoc, Dipodascus, Lactococcus lacti, Dipodascus australiensis) are gradually increased in concentration, and the potentially detrimental microbial genus and species (e.g., Acinetobacter, Pseudomonas, Fusarium, Aspergillus, Mortierella, Acinetobacter johnsonii, Fusarium solani) decrease during the decline of bacterial and fungi diversity from natural fermentation. The study of microbial community dynamics could make a great contribution to understand the mechanism of natural fermentation of milk and the formation of curds and Zuohe, and to discover the potentially fermentative microbes for industrial starter cultures.

from increased metabolite production by fermentative bacteria and fungi.
However, few studies addressed the completely natural static fermentation, the intermediate process for the production of cheese and Zuohe (traditional sour cream). The sour cream moves upwards F I G U R E 1 Schematic representation of static, natural fermentation of milk. 100 kg of raw cow's milk was naturally fermented at room temperature (6.3 to 16.7℃), and microbial community dynamics (44 samples) was investigated by using culture-dependent methods and high-throughput amplicon sequencing F I G U R E 2 Microbial dynamics during natural fermentation of milk. LAB, yeast, and mold of samples from the continuous fermentation process were quantified. The dynamics of this process included four rise and fall stages of LAB (L1, L2, L3, and L4), four fall and rise stages of yeast (Y1, Y2, Y3, and Y4), and one fall and rise phase of mold (M4)  (Akabanda et al., 2013;Gao et al., 2017;Liu et al., 2015;Mathara et al., 2004;Nahidul-Islam et al., 2018;Oki et al., 2014;Shangpliang et al., 2017Shangpliang et al., , 2018Sun et al., 2014;Takeda et al., 2011;Yamei et al., 2019;Yao et al., 2017;Yu et al., 2011), neglecting the dynamics of the microbial community during natural fermentation.

TA B L E 1 Microbial changes during natural fermentation of milk
In this study, the microbial community, including LAB, yeast, and mold, during the process of natural fermentation of cow's milk were investigated by using culture-dependent methods and highthroughput amplicon sequencing including 16S rRNA and internal transcribed spacer (ITS) to understand the detailed fermentation process, the roles of distinct fermentative microorganisms, and the interplay between bacteria and fungi.

| Construction of the Natural Fermentation Model and Quantification of LAB, Yeast, and Mold
To simulate natural milk fermentation, 100 kg of raw milk of Holstein from Xilingol prairie was used to ferment at low temperature (6.3 to 16.7℃) (Figure 1). We hypothesized that the fermentation of large amounts of raw milk could stabilize microorganism dynamics, which makes the results more representative. Large-scale, low temperature natural fermentation was established for investigating bacterial and fungal community dynamics in details, which helps us further understand the changes and relationships of LAB, yeast, and mold in the process of natural fermentation. The protein, fat, and lactose content of raw cow's milk was 3.27%, 3.88%, and 4.92%, and the acidity and pH were 15°T and 6.85. We maintained the temperature of fermentation below 20℃ to extend the period of natural fer-  (Edgar, 2013). The OTU was classified into organisms by the Naive Bayesian Model using RDP classifier (Wang et al., 2007) based on SILVA database for 16S rRNA sequencing (Pruesse et al., 2007) and UNITE database for ITS sequencing (Koljalg et al., 2005).

| Viability changes in LAB, yeast, and mold during natural fermentation of milk
The total number of LAB and yeast in raw milk at the beginning of the fermentation process were 6.22 log cfu/ml and 4.16 log cfu/ml, respectively. In contrast, we did not detect any mold at this stage of the process. Given the small changes in the microbial counts, as well as in the pH, in early stages of fermentation, samples were collected every 8 hr followed by every 4 hr in middle and later stages. As shown in Figure 2 and Table 1, total LAB number significantly increased from 0 hr (6.22 log cfu/ml) to 32 hr (9.57 log cfu/ ml) to 44 hr (9.86 log cfu/ml) to 56 hr (10.83 log cfu/ml) to 80 hr (11.98 log cfu/ml) to 92 hr (12.29 log cfu/ml). Total yeast number rapidly increased after 24 hr (6.80 log cfu/ml), then gradually decreased until 96 hr (4.56 log cfu/ml). After 68 hr of fermentation, we detected the presence of mold (2.7 log cfu/ml), which gradually increased until 96 hr (4.48 log cfu/ml) ( Figure 2). We further observed a decrease in the pH from 6.85 at the beginning of fermentation, to 4.66 at 96 hr. After 88 hr of fermentation, we observed floating sour cream (pH 4.82). We collected the superficial sour cream (Z1, Z2, and Z3) as well as underlying curds at 88 hr, 92 hr, and 96 hr, respectively. The average number of viable LAB, yeast, and mold in the Zuohe was 12.18 ± 0.50 log cfu/ml, 6.05 ± 0.30 log cfu/ml and 5.41 ± 0.24 log cfu/ml, respectively ( Figure 2 and 12.18 ± 0.50 log cfu/ml) were higher than in any of the above fermented products tests. In addition, the viable yeast count ranged from 2.41 to 6.98 log cfu/ml in Chigee (Guo, Xu, et al., (Witthuhn et al., 2004). In contrast, yeast counts in our naturally fermented products (the curds: 4.15 ± 0.38 cfu/ml; Zuohe: 6.05 ± 0.30 cfu/ml) were generally lower. We reasoned that the large-scale, low temperature, and static natural fermentation could contribute to LAB proliferation, which in turn might inhibit the yeast growth. We hypothesize that LAB and yeast in coculture may compete for nutrients or that they produce some metabolic substances that inhibit each other's growth.

| Dynamic change of bacterial community during consecutive natural fermentation of milk
After removing low-quality and chimera reads, a total of 2,133,019 bacterial reads (Average ± SD: 88,876 ± 4,210) were obtained, and the OTUs of 0 hr to Z3 are shown in Table 2. Chao1, Shannon, Simpson, and Good's coverage were utilized to evaluate bacterial community enrichment and diversity, and shown in Table 2. These indexes demonstrated that the bacterial community was adequately represented, and changes of OTUs and Shannon indicated that bacterial diversity declined during natural fermentation. In addition, there were no significant differences in bacterial diversity between the curds (88 hr, 92 hr, and 96 hr) and Zuohe (Z1, Z2, and Z3) (p > .05).

| Bacterial community variations in the representative fermentation stage
To assess the bacterial dynamics during natural fermentation, some were obtained, and the OTUs are shown in Table 3. Chao1, Shannon, Simpson, and Good's coverage were used to evaluate bacterial community enrichment and diversity, as shown in Table 3. The alpha indexes demonstrated that the bacterial diversity declined during natural fermentation (p < .01), with higher bacterial diversity in the Zuohe than in the final curds (96 hr, p < .01).

F I G U R E 5 Relative abundance of fungal sequences (genus level) in the samples from the continuous fermentation process (a). The
representative genera change during the continuous three stages of fermentation (b). The three consecutive sampling time points can be classified as a fermentative stage, namely as S1 to S7, and the superficial sour cream (Z1, Z2, and Z3) were collected at 88 hr, 92 hr, and 96 hr declined from 0 hr to 96 hr (Table 4). In addition, the Zuohe exhibited higher abundance of Proteobacteria than the curds, and less Firmicutes (p < .01) ( Table 4) and Pseudomonas declined significantly (p < .01). Streptococcus increase, after an initial decrease in abundance (Figure 4c and Table 4)

TA B L E 5
Fungal diversity indices of ITS sequencing of the samples from the continuous fermentation process (p < .01), whereas Obesumbacterium initially increased, followed by a significant decrease (Figure 4c and Table 4) (p < .01). The abundances of Lactococcus and Leuconostoc were lower in the Zuohe than in the curds (p < .01). In contrast, Acinetobacter, Pseudomonas, Streptococcus, and Obesumbacterium were more present in the Zuohe than in the curds (p < .05) (Table 4). Regarding the bacterial species group, Lactococcus lactis showed a significant increase (p < .01), contrary to Acinetobacter johnsonii, which abundance declined significantly (p < .01) (Table 4). Furthermore, the amount of Lactococcus lactis in the Zuohe was lower than in the curds (p < .01), whereas Acinetobacter johnsonii was higher (p < .05) ( Table 4).
Each of the previously mentioned bacteria as a key role during natural fermentation. Lactococcus strains are widely used for industrial production of fermented dairy products. Besides the capacity to extend the shelf-life of dairy products, increasing amounts of Lactococcus during milk natural fermentation are at the basis of the sour and fermentative fragrances (Casalta & Montel, 2008;Cavanagh et al., 2015;Song et al., 2017). Leuconostoc spp. are a major contributor to the production of aromatic compounds during dairy fermentations (Endo et al., 2020). Streptococcus thermophilus is a species of lactic acid bacteria which is essential for the manufacturing of many types of fermented dairy products (Harnett et al., 2020). Acinetobacter (Kämpfer, 2014), Pseudomonas (Dodd, 2014), and Obesumbacterium (Enterobacteriaceae family) (Patel et al., 2014) are regarded as spoilage microbes for food, bringing about concerns human health.
Multivariate analysis was performed to compare the bacterial community structures from naturally fermented samples. As demonstrated in Figure 7a, PcoA, which uses species-level OTUs, showed significant differences among samples from different fermentation time points (ANOSIM, R = 0.82, p = .001), supporting the successional dynamics of bacteria. In addition, samples from different fermentation time points were largely separated in the bray analysis (accounting for 60.16% and 19.33% of the total variance by the two principal components, respectively). In conclusion, the process of natural fermentation is accompanied by the growth of viable LAB count and the decay of bacterial diversity. At the same time, dairy fermentative microorganisms gradually increase, whereas potential spoilage or pathogenic microbes decrease dramatically.

| Dynamic change of fungal community during consecutive natural fermentation of milk
After removal of the low-quality and chimera reads, a total of 1,856,760 fungal reads (Average ± SD: 77,365 ± 10,108) were obtained. OTUs of 0 hr to Z3 are shown in Table 5. Chao1, Shannon, Simpson, and Good's coverage were used to evaluate fungal community enrichment and diversity, as shown in  (Figure 5a).

| Fungal community variations in the representative fermentation stage
To assess fungal dynamics during natural fermentation, key fermen-  Table 6. Chao1, Shannon, Simpson, and Good's coverage were used to evaluate fungal community enrichment and diversity, as shown in Table 6. The alpha indexes showed that fungal diversity declined during natural fermentation (p < .01), with no significant differences in the fungal diversity between the curds (96 hr) and Zuohe (Z3) (p > .05). At the phylum level, the phyla Ascomycota (p < .01) increased during natural fermentation, whereas Mortierellomycota (p < .01) and Basidiomycota (p < .01) declined from 0 hr to 96 hr (Table 7). At the genus level ( Figure 6 and Table 7), Dipodascus increased significantly (p < .01), and Aspergillus, Fusarium, and Mortierella showed an initial increased (0 hr to 44 hr) followed by a marked reduction (Figure 6b and Table 7) (p < .01). Concerning fungal species group, Dipodascus australiensis increased significantly (p < .01), which contrasts with Fusarium solani, which declined after an initial increase (from 0 hr to 36 hr, p < .01) ( Table 7). We did not observe significant differences in terms of the abundances of fungal phylum, genus, and species between the curds and Zuohe (Table 7) (p > .05).
The yeast Dipodascus has been identified in naturally fermented dairy products from Inner Mongolia Yamei et al., 2019). Aspergillus, Fusarium, and Mortierella present potential threats to cattle and humans that might result from fungal abortion, mycotoxins, and aspergillosis (Davies et al., 2010;Foster, 2017;Rodrigues, 2016;Thrane, 2014). We reasoned that, in the context of dairy products, Aspergillus, Fusarium, and Mortierella behave as F I G U R E 6 Relative abundance of fungal sequences (genus level) in the representative samples of fermentation (a). The representative genera change during the representative stages of fermentation (b) spoilage microorganisms. After the initial growth (0 hr to 44 hr), these three genera become feeble and gradually disappear. Raw milk is a favorable environment that supports their initial growth.
However, growth is likely inhibited by the increasing amounts of LAB and, consequently, of increased lactic acid concentration. This might inhibit the growth of these fungi in the stages of metaphase and anaphase of natural fermentation. A reduction in potentially pathogenic molds throughout the natural fermentation meant that potential mycotoxins were produced in the natural fermentation of cow's milk. As a result, further studies will be carried out for the detection and quantification of mycotoxins in the traditionally fermented milk.
Multivariate analysis was performed to compare the fungal community structures from samples during natural fermentation. As demonstrated in Figure 7b, PcoA, which uses species-level OTUs,

| CON CLUS IONS
Traditional dairy products derived from milk's natural fermentation are worldwide used for their nutrient content, fermented flavor and long shelf-life, providing substantial benefits for human health.
Still, the presence of potentially pathogenic microorganisms derived from natural fermentation, together with lack of proper sanitary conditions raised public concerns. In this study, we found that the beneficial bacteria and fungi (e.g., Lactococcus, Streptococcus, Leuconostoc, Dipodascus) are gradually increased in concentration and that potentially pathogenic microorganisms (e.g., Acinetobacter, Pseudomonas, Fusarium, Aspergillus, Mortierella) decrease during the process of natural fermentation. The results support the health beneficial properties of naturally fermented products and highlight the nomadic dairy culture for consumption of naturally fermented milk. Natural, static milk fermentation forms curds and Zuohe (traditional sour cream). Although there were no significant differences in the cultured LAB between the underlying curds and superficial Zuohe, the potentially detrimental bacteria (Acinetobacter, Pseudomonas, Fusarium) were significantly increased in the second, which results in an increase of its bacterial diversity. Furthermore, the Zuohe was significantly more enriched in yeast and mold than the curds. However, there were no significant differences in fungi diversity in terms of genus and species between the two types of fermented samples. Given that the Zuohe was at the surface, and thus exposed to environmental microorganisms, it was expected that the potentially detrimental bacteria, yeast, and mold in the air were also significantly more enriched in the Zuohe than in the underlying curds.

CO N FLI C T O F I NTE R E S T
All authors declare no conflict of interest.