Metagenomic analysis and antimicrobial activity of two fermented milk kefir samples

Abstract In recent years, the fermented milk product kefir has been intensively studied because of its health benefits. Here, we evaluated the microbial consortia of two kefir samples, from Escarcega, Campeche, and Campeche (México). We considered a functional comparison between both samples, including fungal and bacterial inhibition; second, we applied shotgun metagenomics to assess the structure and functional diversity of the communities of microorganisms. These two samples exhibited antagonisms against bacterial and fungal pathogens. Bioactive polyketides and nonribosomal peptides were identified by LC‐HRMS analysis. We also observed a high bacterial diversity and an abundance of Actinobacteria in both kefir samples, and a greater abundance of Saccharomyces species in kefir of Escarcega than in the Campeche kefir. When the prophage compositions were evaluated, the Campeche sample showed a higher diversity of prophage sequences. In Escarcega, we observed a prevalence of prophage families that infect Enterobacteria and Lactobacillus. The sequences associated with secondary metabolites, such as plipastatin, fengycin, and bacillaene, and also bacteriocins like helveticin and zoocin, were also found in different proportions, with greater diversity in the Escarcega sample. The analyses described in this work open the opportunity to understand the microbial diversity in kefir samples from two distant localities.


| INTRODUC TI ON
The use of fermented products has been known to humanity for centuries, and the search for their health benefits has increased in the last decade through research focused on various food products.
One of these products of increased interest is kefir. Kefir has been associated with health benefits, such as reduction of blood pressure (Klippel et al., 2016), immunoregulation (De Montijo-Prieto et al., 2015), and antiallergic, antitumoral, antimicrobial, and antiinflammatory effects (Arena et al., 2019;De Montijo-Prieto et al., 2015;Gao & Zhang, 2019;Hong et al., 2010;Seo et al., 2018;Sharifi et al., 2017) among other health benefits. Kefir is a fermented milk product that is an aggregate of microorganisms, in which lactic acid bacteria, acetic acid bacteria, and yeasts have been reported as predominant (Pogačić et al., 2013;Zhou et al., 2009). However, the microflora and the predominant bacteria in kefir may vary depending on the substrate used in the fermentation process, the method of maintaining the culture, and the geographical, climatic, and cultural conditions, as well as the type of milk used (Marsh et al., 2013;Prado et al., 2015).
There is evidence indicating that microorganisms from a kefir's consortium produce several metabolites, including phosphopeptides, peptides, antibiotics, exopolysaccharides, and bacteriocins, that inhibit the development of degrading microorganisms and pathogens, such as Salmonella, Helicobacter, Shigella, and Staphylococcus (Anton et al., 2016;Cleveland et al., 2001;Hong et al., 2010;Lopitz-Otsoa et al., 2006;Prado et al., 2015). Kefir can also inhibit pathogenic fungi, such as Candida albicans and Fusarium graminearum CZ1, among others (Ismaiel et al., 2011;Lopitz-Otsoa et al., 2006), and it inhibits Aspergillus flavus formation of spores and production of aflatoxin B1, a toxic compound formed in the crop field or during food storage (Ismaiel et al., 2011). In addition, a high proportion of volatile organic and aromatic compounds in kefir, such as lactic acid, acetic acid, and butyric acid, as well as ethanol, have also been described as important in the inhibition of fungi and bacterial growth (Cais-Sokolińska et al., 2015;Magalhães et al., 2011). Therefore, considering the importance of kefir in diverse health and antimicrobial mechanisms, we carried out a systematic study of two kefirs, from Campeche (C_kefir) and Escarcega (E_kefir), by an exhaustive functional analysis describing the antagonistic effects against bacterial and fungal pathogens, complemented with a metabolomic profile compiled from liquid chromatography-highresolution mass spectrometry (LC-HRMS) data to identify probable bioactive compounds involved. Besides, the microorganism and functional diversities were determined for these samples, using a metagenomic shotgun approach. We consider that this analysis opens diverse opportunities to understand the functional role of the microbial consortia, microbial diversity, and their functional profiles within kefir lactic fermented beverage systems and will contribute to knowledge about these environments.
The second kefir sample, Escarcega's kefir (E_kefir), was collected from a cattle farm in the municipality of Escarcega, Campeche, Mexico (latitude 18.617, longitude −90.717; 18° 37′ 1″ north, 90° 43′ 1″ west). Both kefir granule samples were kept in the milk of commercial origin, to ensure that the quality of the milk would always be as homogeneous as possible and to diminish significant changes in the microbial consortia during storage.

| Titratable acidity
Titratable acidity was determined from fermented milk after 48 hours. To 20 ml of filtered kefir, 3 drops of phenolphthalein were added, after which the mixture was titrated with sodium hydroxide, with the observation of the spent volume until the milk maintained its pink coloration for more than 30 s. It was released three times.
The percentages of total organic acids were calculated as follows

| Production and detection of compounds with inhibitory activity
To evaluate the production of antimicrobial metabolites produced by the lactic acid bacteria, consortia were cultured in de Man, Rogosa, and Sharpe (MRS) medium and incubated for 24 h at 25°C. Subsequently, the cultures were centrifuged at 2800 g for 20 minutes at 4°C, and the supernatant was filtered through a membrane of 0.22 µm diameter. These extracts were tested against phytopathogenic fungi (Curvularia sp., Fusarium equiseti, and Colletotrichum gloeosporioides), and against human commensal and pathogenic bacteria (Escherichia coli, Salmonella typhimurium, Bacillus subtilis, and Staphylococcus aureus), using the method of Kirby-Bauer for bacteria and radial inhibition percentages for fungi (Bauer et al., 1966).

| Antifungal activity
We evaluated the activity of kefir against the growth of three patho- where Rc is the mean value of the fungus radius in the absence of kefir and Ri represents the fungus radius in the presence of the consortium.

| Chromatographic analysis
Sample separation was performed in a Biozen C18 RE XB column (100 × 21 mm, 1.7 mm particle size), with water and acetonitrile as mobile phases A and B, respectively. Both phases were used in a gradient mode with a flow of 0.2 ml/min: 0% B for 3 min, 15% B for 5 min, followed by 35% in 5 min, increased to 100% in 7 min, leaving 100% B for 4 min, returning to 100% A in 4 min, finishing in 30 min of analysis. For the analysis, 5 μl was injected. After separating, peaks were passed to the MS detector.

| Bioactive metabolite identification
The identity of compounds in the extracts was determined by searching the monoisotopic accurate mass generated during the analysis in the MetLin commercial database implemented in the Agilent Masshunter program integrated with the PCLD software.
MetLin comprises 29,000 exogenous and endogenous natural products from diverse sources, including actinomycetes. Alternatively, an in-house Bacillus and Gordonia bioactive metabolite database was built after retrieving data reported in the literature (Farzand et al., 2019;Nagao et al., 2001;Pan et al., 2019;Phister et al., 2004;Xu et al., 2018), using the PCLD program in the Masshunter program (Agilent Technologies). Special care was taken with the sodium and potassium adducts that sometimes form in an MS analysis.

| DNA extraction and sequencing
The genomic DNA of C_kefir and E_kefir was extracted using the MoBio Power Soil kit. The samples obtained were sequenced using the Illumina platform (Illumina, 2017

| Taxonomic annotation
The quality of reads was assessed by using FASTQC v0.11.4 software (Andrew, 2010). Subsequently, the reads were trimmed and adapters were removed using Trimmomatic software v.0.38. The reads were assembled using Megahit v 1.1.1-2, with a -k min = 21, k max = 99. The

Radial inhibition
taxonomic annotation was performed using two strategies: In the first one, the assembly contigs were annotated by Kaiju (Menzel et al., 2016); alternatively, a second strategy the Metagenomic Rapid Annotations server using Subsystems Technology (MG-RAST) was considered.
The diversity index was evaluated using alpha and beta descriptors within the Phyloseq library, and sampling effort was evaluated through the rarefaction curves using a Vegan library implemented in R.

| Prophage sequence search in metagenomes
For the identification of prophage, the first approach was through a comparison in BLAST, using a complete genome database of viruses, containing 6000 genomes, with the following parameters: the number of alignments = 20, e-value = 0.0001, and word size = 11.
Subsequently, MEGAN was used to perform the taxonomic classification based on the lowest common ancestor (LCA) and the parameters minimum support = 2, minimum score = 70, top percent = 10.
The second strategy that was used was through VIBRANT standalone (Kieft et al., 2020). Briefly, using the hybrid machine learning and similarity of proteins approach to recover the complete virus, the parameters used were contigs with a minimum length of 1000 bp, summary plots on, and function virome off, and the ORF number per scaffold was set to 4 to limit the input to sequences.

| Morphological and physicochemical features of kefir granules
The kefir granules were collected from two locations in the southeast of Mexico, Campeche (C_kefir) and Escarcega (E_kefir). The granules of both kefirs exhibited a similar lobular and irregular shape; C_kefir has granules of 2-4 mm in diameter that are milky white with a firm and viscous texture. In counterpart, granules of E_kefir have a size of 1-2 mm in diameter and a creamy color. These morphological characteristics have been also reported for Argentinean and Tibetan kefirs (Chen et al., 2015;Garrote et al., 2001). At 48 h, both consortia presented a white creamy and carbonated consistency; the pH was 3.7 and 3.6 for C_kefir and E_kefir, respectively, that is, they were not significantly different, as already reported after 48-h incubation (Garrote et al., 2001).
In addition, the titratable acidity observed in C_kefir was 0.733 g/L, while for E_kefir, it was 0.792 g/L. It is known that organic acids are the major end products of milk fermentation at 48 hours and are associated with a pH decrease, making an acidic environment (Garrote et al., 2001;Sung-Ho et al., 2013). These organic acids are also associated with the organoleptic and antagonistic properties of kefir (Bengoa et al., 2019). In summary, morphological characteristics and physicochemical properties revealed that pH and titratable acidity were similar between the two kefir samples and were consistent with previous descriptions (Garrote et al., 2001;Hong et al., 2019;Sung-Ho et al., 2013).

| Kefir exhibits an antagonistic effect against fungal pathogens
It has been described that kefir inhibits pathogenic fungi, such as C. albicans, F. graminearum CZ1, and A. flavus. Therefore, to determine whether both Campeche and Escarcega kefirs exhibited antifungal activities, suspensions and cell-free extracts were evaluated.
The first experiment considered a total suspension of both kefirs in a dual-culture antagonism assay on PDA plates with three phytopathogenic fungi, Curvularia sp., Fusarium equiseti, and Colletotrichum gloeosporioides. These fungi were selected because they have a wide spectrum of hosts and cause great loss of crops in Mexico. The inhibitory activity was determined based on the limited growth of fungal mycelia in the inhibition zone. In Figure 1, we show that the highest inhibition corresponds to C. gloeosporioides with 71% radial inhibition with the E_kefir and 59% radial inhibition with C_kefir.
For Curvularia sp., the inhibition was 68% with E_kefir and 56% with C_kefir, whereas for F. equiseti, it was 50% and 40% with C_kefir and E_kefir, respectively. These results indicate that both kefir suspensions inhibit significantly the three phytopathogenic fungi tested in this assay.
To determine whether the inhibition observed with the two consortia was due to compounds secreted by the microbial community tabolites that inhibit A. fumigatus and A. nidulans (Lind et al., 2005).
It has also been reported that the antifungal activity of 91 isolates of lactic acid bacteria was attributed to the presence of lactic, acetic, and phenylacetic acids and by a peptide produced by Lactobacillus fermentum (formally Limosilactobacillus fermentum) (Gerez et al., 2013;Zheng et al., 2020). Similarly, the antifungal activity can be attributed to the synergistic effect between all the organic acids of the fermentation and by antimicrobial peptides (Arena et al., 2019). In summary, our results suggest that the inhibition of kefir is the result of not only molecules secreted by the microbiota but also the competition for the niche and/or for nutrients, as the inhibition observed with cell-free extracts was less extensive than the antagonistic effect by the total kefir extracts. In this regard, the main mechanism of inhibition of lactic acid bacteria is by a synergistic effect between the metabolites secreted and the competition for niche and nutri-

| Antibacterial activity by cell-free extracts of kefir
To determine whether the extracts with antifungal activity (described above) also exhibited antibacterial activity, the cell-free extracts were challenged against two pathogenic bacterial strains, S. typhimurium ATCC 14028 and S. aureus WT, and two commensal bacterial strains, E. coli MG1655 and B. subtilis ATCC 23857. From this assay, we found that the C_kefir and E_kefir extracts inhibited the four bacterial strains, at no dilution, dilution of 1:2, and dilution of 1:4 ( Figure 3). Also, the E_kefir extracts showed bactericidal activity against the four bacterial strains with no dilution and a dilution of 1:2, considering that no growth of colonies was observed in plates of culture medium seeded in incubation after 24 h. Therefore, C_kefir cell-free extracts showed increased bactericidal activity against the four bacterial strains, in comparison with the E_kefir extracts, suggesting that C_kefir is more efficient in inhibiting bacterial growth.
In this address, antagonistic effects of lactic acid bacteria against Therefore, the production of some inhibitory compounds, such as bacteriocins, hydrogen peroxide, and organic acids, might be responsible for killing pathogenic microorganisms (Silva et al., 2009). Our results suggest that both cell-free extracts from C_kefir and E_kefir have antifungal and antibacterial activities, probably related to the production of compounds secreted by the microbiota that conform to both kefirs.

| Metabolomic profile by nontargeted LC-HRMS
To identify the chemical nature of the compounds, present in the cell-free extracts of kefir, liquid-liquid organic extraction with chloroform was performed, and metabolomic profiles for both samples were obtained. We used a solvent extraction method because we were interested in amphiphilic molecules, such as bacteriocins produced by lactic acid bacteria. In this regard, hydrophobic regions in antimicrobial molecules are central in their affinity to the lipidic membrane of the cell (Abee et al., 1991;Yusuf, 2013). The analysis showed that E_kefir presents more signals in the chromatogram than the C_kefir consortium. ( Figures   A1 and A2). Based on a nontarget LC-HRMS study, we identified 11 different bioactive compounds between the two consortia based on the accurate monoisotopic molecular weight (Table 1 and Figure A3). To do this, we used an in-house database with F I G U R E 1 Radial growth inhibition shows the antagonistic effect of the total kefir extracts against three fungi. Columns are as follows: C. gloeosporioides. (Column 1), Curvularia sp. (Column 2), and F. equiseti (Column 3). In lines are the E_kefir (Line 1), C_kefir (Line 2), and Control, fungi growing in PDA medium with no extract (Line 3). The % of radial inhibition is shown. n = 3 Bacilysin, bacillaene, and macrolactins are polyketides that belong to a large class of structurally diverse natural products that exhibit an extensive set of biological activities, such as antimicrobial activities (Chan et al., 2009;Hill et al., 2017;Park et al., 2017;Schneider et al., 2007). Although the metabolomic profile only differs from two metabolites, kammogenin is lacking in C_kefir and Macrolatin H is presented only in the C_kefir. Indeed, C_kefir showed more concentrated metabolites. This result is consistent with the previous analysis, where compounds produced by kefir exhibit different spectra and activities according to the fermentation time (Kim et al., 2016).
Based on these results, open questions remain to be explored: What is the microbial composition of C_ and E_kefirs? Do bacterial consortia produce different compounds associated with their microbial population? Therefore, in the following sections, we describe our main findings associated with a metagenomic analysis to determine the diversity, abundance, and metabolic profiles associated with both kefirs.

| Kefir is a consortium integrated by a large proportion of bacteria and Eukarya organisms
To determine the organisms associated with the production of compounds previously described, the microbial and metabolic di-

The presence of Bacteria and Eukarya at different proportions
suggests that their contribution could influence the production of more bioactive compounds in the E_kefir and C_kefir. For instance, as reported in water kefir, the interaction in coculture between L. kefiranofaciens and S. cerevisiae enhances the production of kefiran a polysaccharide with antimicrobial activity (Cheirsilp et al., 2003).
When the metagenomes were analyzed at the genus level,  (Marsh et al., 2013).
The indices of richness and evenness were calculated, and the results indicate that the diversity of E_kefir is much greater than the diversity of C_kefir (Table A1). Also, a similar trend can be observed in the rarefaction curve ( Figure A5), where E_kefir was close to reaching a horizontal asymptote, compared to C_kefir. These results indicate that E_kefir in general has a more diverse consortium than C_kefir.

F I G U R E 5 Heatmap of the taxonomic classification of recovered bacteriophage contigs. (a) Family, (b) Species
These variabilities in the populations could be associated with the production of different compounds.

| Prophage diversity in kefir metagenomes
Bacteriophages play a pivotal role in microbial abundance and metabolism, due to their ability to regulate the competitive relationships among different microorganisms (Mills et al., 2013). To determine the diversity of prophages, we retrieved those prophage sequences from the metagenomic DNA described above. From this analysis, we found in C_kefir 0.19% of the sequences corresponded to prophage sequences, versus 0.25% in E_kefir. According to our results, we found that C_kefir showed a greater diversity of prophages than did E_kefir, and we observed a prevalence of fami-

| Prediction of secondary metabolites produced by C_kefir and E_kefir
To identify probable genes encoding the biosynthetic pathway for the production of secondary metabolites in the metagenomic sequences of C_ and E_kefirs, the program antiSMASH (Blin et al., 2019) was used. In brief, antiSMASH uses a collection of profiles to predict clusters of genes associated with secondary metabolite biosynthesis pathways. Based on this approach, we identified 18 putative biosynthetic gene clusters in C_kefir and 40 in E_kefir that are responsible for the production of secondary metabolites. These clusters of genes were identified as associated with the production of bacteriocins, polyketides (PKs), and nonribosomal peptides (NRPs), active against a wide range of microorganisms including bacteria, protozoa, yeast fungi, prophages, and even tumor cells, in both kefir samples. In this context, in C_kefir we found 14 out of 18 regions associated with the production of NRPs. These regions were identified with a coverage of 44.5 to 100% (Table A2). These napyradiomycin, associated with antifungal and antimicrobial activities, were also predicted. All these compounds are related to Actinobacteria and Bacillales.
In contrast, in the E_kefir samples, we found 40 regions involved in secondary metabolite biosynthesis pathways, according to the antiSMASH program. From these, 32 out of 40 regions are predicted as NRPs; 2 were predicted as PKs and 4 as bacteriocins (Table A3). From the predicted NRPs, 18 were identified as having probable antibiotic effects (coverage of 23.7%-100%), such as nogabecin, plipastatin, daptomycin, macrotermycins, griseoviridin, vancomycin, and virginiamycin, among others, mainly associated with Actinobacteria (Streptomyces) and Bacillales (Bacillus and Paenibacillus). Indeed, this finding correlates with the fact that E_kefir has a greater proportion of Saccharomycetes than C_kefir.
S. cerevisiae has been shown to adjust its metabolism to secrete various metabolites, especially amino acids, which allow the survival of lactic acid bacteria (Ponomarova et al., 2017), and amino acids are the main components of NRP and PK scaffolds.
On the other hand, there is a correlation between results observed by LC-HRMS and antiSMASH. We detected plipastatin in C_kefir extracts by antiSMASH and by LC-HRMS. Also, we detected in E_kefir extracts difficidin, bacillaene, and plipastatin by LC-HRMS and antiSMASH. There is a correlation between the highest antimicrobial activity with E_kefir extracts compared with C-Kefir extracts, agreeing with our results for inhibition, suggesting that E_kefir produces more bioactive secondary metabolites than C_kefir.
The presence of secondary metabolites could explain the antifungal and antibacterial activities of the extracts of both consortia. In this regard, the second group of compounds was predicted, the bacteriocins. Based on an analysis using the BACTIBASE server (Hammami et al., 2010), we found 9 bacteriocins in Campeche and 10 associated with Escarcega (Table A4 and Table A5, respectively).
From these, five bacteriocins classified as zoocin A were predicted in Campeche and seven in Escarcega. Zoocin A has been described as a penicillin-binding protein and presumably is a D-alanyl endopeptidase, identified in several Streptococcus species (Heath et al., 2004).

| CON CLUS IONS
In this work, we studied two kefir samples, from Escarcega and Campeche (México), by two approaches. The first approach was a functional comparison between both samples, including fungal and bacterial inhibition; the second approach used a metagenomic shotgun methodology to assess the structures and functional diversity of the communities of microorganisms. Based on these approaches, we found that these two samples exhibited antagonisms against bacterial and fungal pathogens. Bioactive polyketides (bacillaene, macrolactins, and kammogenin) and nonribosomal peptides (bacilysin, bacillibactin A) were identified by LC-HRMS analysis.
In addition, we observed high bacterial diversity, an abundance of Actinobacteria, and a differential proportion of Ascomycota organisms and prophages. The analyses described in this work provide the opportunity to understand the microbial diversity in kefir samples from two distant localities.

E TH I C S S TATEM ENT
None required.

ACK N OWLED G M ENTS
This study was funded by Tecnológico Nacional de México/IT

CO N FLI C T O F I NTE R E S T
None declared.