Bacterial and yeast diversity
Koumiss taste can be classified into three types according to the lactic acid content: strong (0.91–1.08%), moderate (0.73–0.90%), and weak (0.54–0.72% ) . According to this classification, the koumiss in the present study was classified as a weak form from fermented mare’s milk.
Many studies on koumiss bacterial or yeast diversity used culture- or molecular biology-based methods; however, results varied significantly due to differences in the analysis techniques applied, research approaches, and geographic regions [23-26]. In the current study, we applied long-read SMRT sequencing technology (Pac Bio) with high-resolution phylogenetic microbial community profiling  to identify a novel dominant bacterial species, Citrobacter freundii that had not been reported previously; the dominant yeast species were M. caribbica and uncultured Guehomyces. One observation in our study that was inconsistent with previous research was that the abundance of Enterobacter genus was < 1; this may be because the samples were freshly fermented koumiss, whereas Enterobacter was found to be enriched during later fermentation and is related to flavor ripening . In our study, the abundance of Acetobacter genus was lower (1.5 ± 2.13%), and mainly consisted of Acetobacter malorum, Acetobacter pasteurianus, and Acetobacter orientalis. The heterofermentative species included a small amount of Lactobacillus kefiri (0.82 ± 1.86%) and Leuconostoc mesenteroides (0.53 ± 1.45%).
Regarding most diverse bacterial composition between two groups, the mesophilic bacteria L. lactis was absent from Xilingol-rural koumiss samples, there may be a competition relationship between L. lactis and L. helveticus, because L. helveticus utilizes galactose moieties and glucose, whereas L. lactis only utilizes the latter .Regarding interactions among bacterial species, in a previous study, mixed cultures of Lactococcus lactis ssp. lactis and L. raffinolactis stimulated more acid production during skim milk fermentation . One possible reason for this is that L. raffinolactis strains are able to ferment α-galactosides , which are not utilized by L. lactis strains. However, in the current study, although L. lactis and L. raffinolactis showed a trend toward a positive correlation in their abundances, it was not statistically significant. In contrast, the abundances of L. kefiranofaciens and L. raffinolactis were positively correlated. L. kefiranofaciens is rarely isolated from koumiss  but is a key LAB species in kefir grain ecosystems . Citrobacter freundii (C. freundii) is a common component of the gut microbiome of healthy humans, and of soil, and is a widely distributed opportunistic pathogen. C. freundii was found in one sample at an abundance of 11.37% but was present in very small amounts in other samples; thus, the high-abundance sample may have been due to contamination.
Yeasts producing koumiss (alcoholic) can ferment lactose or galactose. In the current study, the lactose-fermenting yeasts were D. anomala and K. marxianus, while the galactose-fermenting yeasts included K. unispora and M. caribbica.
The predominant yeast species, D. anomala, belongs to the genera Dekkera (Brettanomyces), which is known for its ability to spoil alcoholic beverages by producing acetic acid and unpleasant flavors and odors [36, 37]. D. anomala is the teleomorph of Brettanomyces anomalus; it differs morphologically and physiologically from the other species in the genus Dekkera in terms of the formation of blastese and the ability to ferment lactose . D. anomala IGC 5153 exhibited limited ability to use weak acids as its only source of carbon and energy. The acetate carrier was not only inducible, but also subject to glucose repression, similar to the carriers identified in acetic acid–grown cells of yeasts such as S. cerevisiae. D. anomala grew in acetic acid-containing medium with a relatively high total acid content, and over a broad pH range including pH 3.5. D. anomala isolates were obtained from Inner Mongolia Koumiss samples and shown by polymerase chain reaction denaturing gradient gel electrophoresis (PCR-DGGE) to be present in an amount of 7.29% ; this differs from our sequencing results, which showed amounts of 13–83%. In other milk products, D. anomala was isolated from kefir, koumiss, and Shubat [25, 32, 39]. Feta cheese made from pasteurized ewe’s milk contains D. anomala. Because wood is an ecological niche for this strongly fermented species, its presence is likely attributable to the wooden tables used for dry salting the cheese blocks . In the current study, the two samples with the lowest abundance of D. anomala were fermented in crock using a traditional method. This suggests that non-wood containers and stirrers could be used to restrain the overgrowth of this species. In our koumiss samples, D. anomala may have partially contributed to the production of acetic acid, but its correlation with the acetic acid content was negative and nonsignificant (Figure 6). This may be because it can transport acetic acid into cells as a carbon source to produce CO2; this is known as the Crabtree effect and underlies the “make-accumulate-consume” life strategy of organisms such as S. cerevisiae [41, 42]. It is a strongly fermented species that may bloom at the beginning stage of fermentation.
The second most common yeast species was K. unisporus. A previous study found that its dominance increased at higher altitudes , but in koumiss samples from the Xilingol prairie, where latitude does not vary greatly, the abundance varied widely from 0.03 to 60%. K. unisporus has been found in various dairy products. It is a galactose-fermenting yeast species and the principal microorganism responsible for alcoholic fermentation, but is a slow producer of ethanol and performs a clean fermentation in milk and whey. Thiamine is the only vitamin required exogenously for the growth of K. unisporus; therefore, thiamine may be a key factor determining its prevalence . These samples are weakly fermented koumiss, there is a limited amount of galactose in the system, which may also be another factor.
The yeast Shannon diversity index of Xilinhaote-urban samples was significantly higher than that of Xilingol-rural samples, mainly because of the absence of K. marxianus in most Xilingol-rural samples. K. marxianus utilizes lactose to excrete some glucose and galactose into milk and weakly metabolizes citrate and succinic acids to produce ethanol, glycerol, acetic acid, and propionic acid . K. marxianus is also emerging as a new platform for the production of flavor and fragrance compounds ; it is used for the production of phenylethyl alcohol, responsible for the aroma of roses in cheese whey .
M. caribbica, P. BZ159, and uncultured Guehomyces were detected in the koumiss samples, and their presence may be attributable to the non-starter yeast species. The species within the genus Pichia are methylotrophic yeasts . P. BZ 159 has two alcohol oxidase genes, aox A and aox B, which can oxidize alcohols to aldehydes with concomitant production of hydrogen peroxide. Glucose strongly suppresses the expression of both genes .
Koumiss organic acid profiling
For the first time, we systematically identified the organic acids responsible for the sour taste of koumiss. The organic acid composition did not directly reflect the grouping of samples; in fact, it better-reflected the fermentation process. Most organic acids produced by microbes can be divided into two main groups depending on their metabolic origin and the main metabolite sequence of the aerobic organism: those associated with the tricarboxylic acid cycle (TCA) and glycolysis arising from the oxidation of glucose; and those using glucose to produce organic acids. Citric, lactic, and malic acids fall into the first group, whereas acetic acid should be considered a biotransformation of ethanol. Succinic acid is mostly produced using chemical methods. K. unisporus ferments certain monosaccharides to produce succinic acid and acetic acid during ethanol fermentation , where the lactic and acetic acid content is the main indicator of the fermentation dynamic. In the present study, the average acetic acid content was higher than the lactic acid content, which may be the result of weak fermentation; as the fermentation proceeds, lactic acid may become dominant over acetic acid. In the bacterial community, most bacteria are homofermentative species and the less-abundant heterofermentative species, including L. kefiri and Leuconostoc mesenteroides, produce acetic acid in koumiss. The trace amounts of acetic acid may by generated from ethanol by acetic acid bacteria . More importantly, few LAB species are able to co-metabolize citrate with fermentable sugars to produce acetic acid. Specifically, variants of L. lactis (i.e., L. lactis subsp.lactis biovar diacetylactis) are capable of citrate utilization because they possess a plasmid-encoded citrate transporter gene . Our samples may have contained variants of L. lactis. In the yeast community, acetic acid-producing species include D. anomala and K. unisporus.
There were significant lower content of citric acid in Xilingol-rural than Xilinhaote-urban groups. Citric acid is a common constituent of milk; its abundance in mare’s milk is < 2.3 g/kg. Moreover, it is a substrate for certain LAB, which use it to form aromatic substances such as acetoin and diacetyl [52, 53]. In our study, the more sour Xilingol-rural samples had lower citric acid content, which was consistent with the result that its content decreased with fermentation. During fermentation, the citric acid content did not remain steady. First, microbes use glucose to produce citric aacid via the TCA cycle, and then to form aromatic precursors. In addition, citric, malic, and succinic acids are components of the TCA, which together contributed to sourness and bitterness. There are subtle differences in taste: malic acid has a tart, smooth, and long-lasting flavor, whereas citric acid has a sour and bitter aftertaste that lasts only for a short time. Succinic acid has an acidic and salty/bitter taste and is savory in cheese .
Succinic acid is a good indicator of yeast inoculation status. It is the main organic acid produced by yeast and is formed in the glyoxylate cycle via oxidation of isocitrate, as well as in the reductive citric acid cycle . In the current study, succinic acid accumulation correlated with the abundance of uncultured Guehomyces and L. kefiranofaciens.
Koumiss Microbiota: association with taste
In the past, more attention was paid to bacterial as opposed to yeast species, even though both bacteria and yeast contribute to unique koumiss fermentation. The presence of yeast in dairy products contributes to flavor formation via the synthesis of a variety of chemical compounds . The koumiss starter culture only containing a few strains of microbes  may loss many benefits of koumiss consumption. In tradition, a small aliquot of koumiss from end product is retained for use as starters for the next batch, this method was out of accurate control of fermentation. Redundancy and correlation analysis suggest that those species are associated with tastes and organic acids in koumiss. To control excessive sourness, astringency, and bitterness during koumiss production, our data suggest that the amounts of the post-acidifying strains L. helveticus and D. anomala should be reduced, or those of L. lactis and K. unispora increased, in the koumiss starter culture.
Studies of the interaction between bacteria and yeast in fermented foods yield inconsistent results due to differences in strains, substrates and food matrices [26, 56, 58, 59]. During the fermentation of food products, five types of interactions occur between, and may play a role in the consortia of, microbes . However, the putative correlations identified could not be distinguished in terms of whether they were direct or indirect, nor could directionality be established . A study of the interaction between L. lactis and yeast found that L. lactis produced acetic acid when co-cultured with some yeasts; moreover, the growth of L. lactis was enhanced in most co-cultures in UHT milk . The present study observed a potential symbiotic relationship between bacteria and yeast; however, the mechanism underlying this putative relationship is unclear so further research is necessary.