High-throughput sequence-based analysis of the bacterial composition of kefir and an associated kefir grain


  • Editor: Wolfgang Kneifel

Correspondence: Paul D. Cotter, Teagasc Food Research Centre, Moorepark, Fermoy, Co. Cork, Ireland. Tel.: +353 0 2542 694; fax: +353 0 2542 340; e-mail: paul.cotter@teagasc.ie


Lacticin 3147 is a two-peptide broad spectrum lantibiotic produced by Lactococcus lactis DPC3147 shown to inhibit a number of clinically relevant Gram-positive pathogens. Initially isolated from an Irish kefir grain, lacticin 3147 is one of the most extensively studied lantibiotics to date. In this study, the bacterial diversity of the Irish kefir grain from which L. lactis DPC3147 was originally isolated was for the first time investigated using a high-throughput parallel sequencing strategy. A total of 17 416 unique V4 variable regions of the 16S rRNA gene were analysed from both the kefir starter grain and its derivative kefir-fermented milk. Firmicutes (which includes the lactic acid bacteria) was the dominant phylum accounting for >92% of sequences. Within the Firmicutes, dramatic differences in abundance were observed when the starter grain and kefir milk fermentate were compared. The kefir grain-associated bacterial community was largely composed of the Lactobacillaceae family while Streptococcaceae (primarily Lactococcus spp.) was the dominant family within the kefir milk fermentate. Sequencing data confirmed previous findings that the microbiota of kefir milk and the starter grain are quite different while at the same time, establishing that the microbial diversity of the starter grain is not uniform with a greater level of diversity associated with the interior kefir starter grain compared with the exterior.


Kefir is a slightly carbonated fermented beverage manufactured through the fermentation of milk with kefir starter grains. These grains are unique dairy starters that contain a symbiotic consortium of microorganisms strongly influenced by grain origin and culture conditions (Garrote et al., 2010). Although the total number of microorganisms and their relative composition in grains is variable and ill-defined, kefir grains have been shown to contain lactic acid bacteria (LAB; primarily lactobacilli and lactococci), yeasts, and occasionally acetic acid bacteria, within a protein–lipid–polysaccharide solid matrix (Lopitz-Otsoa et al., 2006). The starter grains are vital components for the kefir fermentation as the finished product does not possess the same number or complexity of microorganisms and therefore cannot be used to reinitiate further kefir fermentations (Simova et al., 2002; Farnworth, 2005). Following the fermentation process the kefir grains can be recovered, reused, and grown, often over periods of several decades. In addition to the value of the kefir-associated microbial community as a whole, specific strains isolated from kefir may have value as probiotics (Golowczyc et al., 2008) or as producers of antimicrobial compounds (Ryan et al., 1996; Rodrigues et al., 2005). However, the symbiotic nature of the kefir microbiota can make the identification of such strains and their subsequent investigation more complicated. Although culture-dependent techniques and traditional molecular methods, such as denaturing gradient gel electrophoresis (DGGE) and Sanger sequencing have been used to characterize the kefir community (Ninane et al., 2007; Zhou et al., 2009; da Miguel et al., 2010), such methods may provide an inaccurate description of the total microbial structure in that they reveal only dominant populations, which may not necessarily play important roles in overall community dynamics.

Lacticin 3147 is a potent, two-peptide broad spectrum lantibiotic (class I bacteriocin or antimicrobial peptide) produced by Lactococcus lactis DPC3147 (Fig. 1; Ryan et al., 1996; Martin et al., 2004; Lawton et al., 2007). First isolated from an Irish kefir grain in 1996, it is perhaps one of the most extensively studied bacteriocins and has been shown to inhibit such clinically relevant pathogens as Clostridium difficile, methicillin-resistant Staphylococcus aureus and vancomycin-resistant enterococci (Rea et al., 2007; Piper et al., 2009). Although the microbial composition of kefir grains has been well documented (Rea et al., 1996; Ninane et al., 2007; Zhou et al., 2009), to our knowledge, there have been no reports on the characterization of the microbiota of a kefir grain from which bacteriocin-producing strains have been isolated. In recent years, the field of microbial ecology has been revolutionized by the development and application of high-throughput DNA sequencing technologies, such as that facilitated by the 454 GS-FLX platform (Roche Diagnostics Ltd, West Sussex, UK; Keijser et al., 2008; Urich et al., 2008; McLellan et al., 2009), which allows for a more complete view of overall community composition without the bias typically associated with cloning or cultivation. Here, we use high-throughput sequencing of 16S rRNA gene amplicons to characterize the bacterial composition of the original Irish kefir from which L. lactis DPC3147 was initially isolated.

Figure 1.

 (a) Irish kefir grain. (b) Production of lacticin 3147 by lactococci isolated from an Irish kefir grain as detected by agar well diffusion assay using Lactococcus lactis ssp. cremoris HP as an indicator.

Materials and methods

The kefir grain starter used in this study was obtained from the Teagasc Food Research Centre (Fig. 1a; Teagasc, Fermoy, Ireland) kefir grain collection. The grain was cultured in sterile 10% reconstituted skim milk at 21 °C for 24 h. The fermented kefir milk was removed and the grain rinsed with sterile water to remove any clotted milk still adhered onto the grain surface. In order to monitor bacterial changes over the course of the kefir fermentation, kefir milk samples were enumerated for lactococci and lactobacilli; populations typically associated with the kefir community. Samples were first homogenized as 10-fold serial dilutions, further 10-fold serial dilutions were prepared and appropriate dilutions were spread plated onto M17 agar supplemented with 0.5% lactose (LM17; Difco Laboratories, Detroit, MI) for lactococci, and Lactobacillus selection agar (LBS; Difco) for lactobacilli populations. LM17 plates were incubated aerobically at 30 °C overnight and LBS plates were incubated anaerobically at 37 °C for 5 days. Each experiment was performed in duplicate. Additionally, bifidobacteria were enumerated on modified de Man–Rogosa–Sharpe agar (MRS; Difco) modified with 0.05% cysteine and 100 mg mL−1 mupirocin (Oxoid Ltd, Hampshire, UK). Lacticin 3147 production in the kefir fermentation was also examined using agar well diffusion assays as described previously (Ryan et al., 1996). Briefly, 20 mL of sterile LM17 containing 1.5% (w/v) agar was seeded with 100 μL of the lacticin 3147-sensitive indicator strain, L. lactis ssp. cremoris HP and poured into a sterile Petri dish. Lactococcus lactis ssp. cremoris HP (pMRC01), a lacticin 3147-insensitive derivative of L. lactis HP was also used as an indicator in order to confirm that inhibition of the target strain was solely due to the production of lacticin 3147. Once solidified, wells of uniform diameter were then bored into the medium and 50 μL of the fermented kefir milk was then added to each well. Plates were incubated overnight aerobically at 30 °C and examined for zones of clearing.

For 16S compositional sequencing analysis, genomic DNA from a single kefir fermentation was collected from duplicate samples (∼50 mg) from both the starter grain (interior and exterior surfaces) and kefir milk (2 mL) using the protocols of Garbers et al. (2004) and Lipkin et al. (1993), respectively, and used as a template for PCR amplification of the V4 variable region of the 16S rRNA gene with universal primers [i.e. forward primer F1 (5′-AYTGGGYDTAAAGNG) and R1 (5′-TACCRGGGTHTCTAATCC), R2 (5′-TACCAGAGTATCTAATTC), R3 (5′-CTACDSRGGTMTCTAATC) and R4 (5′-TACNVGGGTATCTAATC)]. Unique molecular identifier (MID) tags were attached between the 454 adaptor sequences and the forward primers. Amplicons generated from two PCR reactions per sample were pooled and cleaned using the AMPure purification system (Beckman Coulter Genomics, Takeley, UK). Sequencing was performed using a 454 Genome Sequencer FLX platform (Roche Diagnostics Ltd) according to 454 protocols. Raw sequencing reads were quality trimmed using the RDP Pyrosequencing Pipeline, applying criteria as outlined previously (Rea et al., 2010). Clustering and statistical analysis of sequence data were performed using the mothur software package (Schloss & Handelsman, 2008). Trimmed FASTA sequences were then subjected to blast analysis using a previously published 16S rDNA gene-specific database and default parameters (Altschul et al., 1997; Urich et al., 2008). blast outputs were parsed using megan (Huson et al., 2007). A bit-score of 86, as previously employed for 16S ribosomal sequence data, was used from within megan for filtering the results before tree construction and summarization (Urich et al., 2008).

Results and discussion

Microbiological analysis

Over the course of the fermentation, lactococci proved to be the dominant microorganism within the kefir fermentation (Fig. 2a). An approximate 5-log increase in presumptive lactococci was observed over the 24 h fermentation period from 7.6 × 104 to 1.1 × 109 CFU mL−1. In contrast, presumptive Lactobacillus counts were lower with an increase of 0.5 log from 3.2 × 105 to 3.7 × 105 CFU mL−1. Only slight changes in the pH of the fermentation were observed during the first 18 h, followed by a sharp decrease to a pH of 4.5 within 24 h. These data are in agreement with previous analyses using Tibetan and Bulgarian kefirs that demonstrated that lactic streptococci, specifically Lactococcus spp., are the dominant microorganisms during the first 24 h of fermentation (Simova et al., 2002; Chen et al., 2008). Furthermore, lacticin 3147 production by L. lactis was detected >8 h into the kefir fermentation and persisted thereafter (Fig. 2b). Indeed, a random sampling of 100 presumptive lactococci isolated from kefir milk (24 h) confirmed that approximately 90% of the colonies analysed were able to inhibit L. lactis HP but not L. lactis HP (pMRC01) indicating them to be producers of lacticin 3147 (data not shown). These results establish that lacticin 3147-producing lactococci are the dominant lactococci present within the kefir-fermented milk. Of particular note with regard to presumptive Lactobacillus populations is the fact that previous culture-dependent analysis of a Turkish kefir found that lactobacilli increased from undetectable levels to 108 CFU mL−1 over the course of a fermentation (22 h) in milk and that similar findings have also been reported with respect to the use of a Brazilian surgary kefir, i.e. lactobacilli increased from 6.6 × 106 to 2 × 108 CFU mL−1 (Magalhaes et al., 2010). As lacticin 3147 has previously been shown to inhibit a number of Lactobacillus species, it is possible that the production of lacticin 3147 may negatively influence cultivable Lactobacillus populations within the kefir community (Ryan et al., 1996). Future work is required to fully elucidate the activity of lacticin 3147 against these populations in this environment.

Figure 2.

 (a) Typical microbial compositional changes during kefir milk fermentation. Lactobacillus spp.; (□), Lactococcus spp.; (◆), pH (χ). (b) Activity (measured as zones of clearing from kefir milk fermentate) recovered from kefir fermentation at indicated time points.

Sequencing analysis

The taxonomic assignments from kefir milk and its corresponding starter grain (interior and exterior) are summarized in Fig. 3. A total of 17 416 unique V4 variable regions of the 16S rRNA gene were amplified from the interior kefir starter grain (4883 reads), exterior starter grain (3455 reads), and the corresponding kefir milk fermentate (9078 reads). Diversity richness, coverage, and evenness estimations were calculated for each data set (Table 1). The Chao1 estimator of species richness at the 98% similarity level was 609.3 for the kefir milk, 170 for the exterior starter grain, and 358 for the interior starter grain. For each sample, the Good's coverage at the 98% similarity level was approximately 98%. A lower level of microbial diversity was observed on the exterior surface of the starter grain with a Shannon diversity index of 1.04 at the 98% similarity level, while the Shannon diversity indices for kefir milk and the interior starter grain were both over 2.0. Rarefaction curve analysis revealed that the overall bacterial diversity present is well represented (Fig. 4). Although the overall bacterial diversity observed within the kefir milk and starter grain were much lower than that of complex ecosystems such as soil or the mammalian gastrointestinal tract (Urich et al., 2008; Turnbaugh et al., 2009), with the exception of the exterior surface of the starter grain, diversity was much higher than that of the only other food system that has been subjected to such statistical analyses, i.e. fermented seafood (Roh et al., 2010). It is possible that inefficient adherence to the kefir grain surface resulting in increased shedding of microorganisms into the kefir milk may be responsible for the decrease in diversity observed on the exterior surface. Alternatively, other microorganisms not identified by 16S compositional analysis (i.e. yeasts) may colonize the majority of the exterior kefir grain surface leading to an underestimation of overall diversity in this region. Notably, previous studies using scanning electron microscopy have revealed that yeast can be densely packed on the exterior of kefir grains (Rea et al., 1996).

Figure 3.

 Family level diversity of microbial communities present in kefir milk, exterior and interior of an Irish kefir starter grain. Bar shading represents the relative proportion of assignable tags present in each sample.

Table 1.   Estimations of diversity within kefir milk and its corresponding starter grain
Data setKefir milkKefir exteriorKefir interior
Chao1 richness estimation396.6609.3124.7170262.4358.4
Shannon's index for diversity1.972.610.831.041.422.1
Good's coverage98.798.298.998.798.498
Figure 4.

 Rarefaction curves for kefir milk and kefir starter grain at the 98% similarity level.

Sequence reads were representative of four different phyla of bacteria, i.e. Firmicutes, Bacteriodetes, Proteobacteria, and Actinobacteria. The Firmicutes were the dominant phylum comprising 92% or more of the total sequences in all samples, while the remaining phyla in combination accounted for just 3.7%, 3.2%, and 0.2% of sequencing reads in the kefir milk, interior starter grain, and exterior starter grain, respectively. It was apparent, however, that the composition of the Firmicutes subpopulation in the milk and grains differed greatly. For instance, a total of 2393 kefir milk-associated reads were assigned to the Lactobacillaceae family (corresponding to 27% of total assigned sequences), while a greater abundance of Lactobacillaceae was observed in the collective starter grain, accounting for 88% (4287 reads) and 96% (3327 reads) of total assignments for the interior and exterior starter grain, respectively (Fig. 3). Although megan outputs could only unambiguously assign sequences of this length to the genus level, further manual investigations of raw blast hits, all with the same bit-score, percentage identity and e-value (scores) allowed the classification of read assignments into a number of Lactobacillus spp. subgroups including Lactobacillus kefiranofaciens, Lactobacillus kefiri, Lactobacillus parabuchneri, Lactobacillus kefiranofaciens ssp. kefirgranum, Lactobacillus helveticus, Lactobacillus acidophilus, and Lactobacillus. parakefiri. Additionally, reads corresponding to the Leuconostocaceae (primarily Leucoconstoc spp.) and the Clostridiaceae families followed similar patterns in that there was an overall greater abundance of taxa assignments corresponding to the interior kefir grain than the exterior or kefir milk. Leuconostocaceae assignments accounted for just 0.1% of assignments in the interior kefir starter grain, but decreased to undetectable levels in the kefir milk and exterior surface. Clostridiaceae assignments accounted for 0.3% of assigned reads in the exterior starter grain, and increased to 0.7% and 0.82% in the kefir milk and interior starter grain, respectively. In contrast, Ruminococcaceae assignments rose from undetectable levels in the kefir grain (both interior and exterior) to 0.1% in the kefir milk. It is possible that local interactions (both antagonistic and symbiotic) that occur between microorganisms in close proximity contribute to the relative differences in the microbial abundances across these two environments (Farnworth & Mainville, 2003).

Conversely, Streptococcaceae (whose members include streptococci and lactococci) assignments comprised just 0.25% of taxa assignments (or 20 reads) in the collective kefir starter grain (including exterior and interior) yet accounted for 65% of assignments (5673 reads) in the kefir milk sample. blast hits, with the same bit-score, included L. lactis, Lactococcus garvieae, as well as uncultured Streptococcaceae and Lactococcus species. The predominance of Streptococcaceae in the kefir milk has been well documented (Rea et al., 1996; Simova et al., 2002; Witthuhn et al., 2005). This is presumably reflective of the Streptococcaceae being more competitive in the milk, relative to the grain, environment as a consequence of their metabolic capabilities and, potentially in this instance, more efficient bacteriocin production. These data confirm previous findings, generated using traditional approaches, that the microbiota of the kefir product and its starter grain can be quite different (Farnworth, 2005). These data are also agreement with our culture-dependent investigations demonstrating the predominance of Lactococcus spp. in the kefir milk (Fig. 2a).

There were a number of notable features with respect to the non-Firmicutes population as well. The Proteobacteria phylum was a minor component of the overall kefir community accounting for just 1.9% of assignments in the interior portion of the starter grain and 0.96% of sequence reads in the kefir milk. Proteobacteria assignments were not detected in the exterior region of the starter grain. Acetic acid bacteria (Proteobacteria subgroup), occasionally associated with the kefir consortium, were not identified within the Irish kefir community, instead Enterobacteriaceae was the dominant bacterial family comprising 1.3% of total assignments in the interior starter grain and 1.67% of reads in the kefir milk. Pseudomonadaceae assignments corresponded to 0.35% and 0.63% of assigned reads in the kefir milk and interior starter grain, respectively. In contrast, Pasteurellaceae represented 0.45% of total assignments in the interior grain but decreased to undetectable levels in the exterior grain and kefir milk. In contrast, Alcaligenaceae rose from undetectable levels in the kefir grain to 0.24% in the kefir milk.

The remaining phyla, Bacteroidetes and Actinobacteria also comprised a minor proportion of the kefir community accounting for a combined 2.73%, 1.3%, and 0.2% of the kefir milk, interior starter grain, and exterior starter grain community, respectively. Of the Bacteroidetes assignments, Bacteriodaceae was the predominant bacterial family with 0.68% of assigned reads in the interior starter grain and 0.8% in the kefir milk (Fig. 3). Bacteroidetes was not detected in the exterior starter grain community. Of the Actinobacteria assignments, Bifidobacteriaceae was the only bacterial family identified in the collective kefir starter grain and kefir milk. To our knowledge, bifidobacteria have not previously been identified as part of the kefir community (Farnworth, 2005; Lopitz-Otsoa et al., 2006). Here the Bifidobacterium population comprised just 0.2% of total taxa assignments in the collective starter grain and 0.4% in kefir milk. blast hits with the same bit-score included Bifidobacterium breve, Bifidobacterium choerinum, Bifidobacterium longum, and Bifidobacterium pseudolongum in both the kefir starter grain and kefir milk. Culture-dependent methods failed to detect Bifidobacterium species in either sample, highlighting the benefits of utilizing a molecular approach. The low percentage of reads corresponding to Bifidobacterium spp. indicates that other molecular approaches, such as DGGE or Sanger-based sequencing, would likely have also failed to detect this subpopulation (Ercolini, 2004). Further studies, involving a number of different grains, are required to establish if members of this generally gastrointestinal tract-associated genus are frequent members of kefir grain populations or if this represents an isolated case.

It is interesting to note that using traditional, culture-dependent approaches, a greater than 1000-fold difference in presumptive Lactococcus (1.1 × 109 CFU mL−1), relative to presumptive Lactobacillus (3.5 × 105 CFU mL−1) populations was observed (Fig. 2a). However, sequencing data established that there is a less than a threefold difference between Streptococcaceae and Lactobacillaceae assignments. This dramatic difference between culture data vs. sequencing results most likely reflects the complex symbiotic relationship observed within the kefir community (Farnworth & Mainville, 2003). It is likely that a number of lactobacilli present within this community cannot be cultivated using standard media and reagents resulting in an inaccurate representation of the overall community.


In this study, the bacterial composition of an Irish kefir grain and its corresponding kefir milk were evaluated using a high-throughput parallel sequencing-based approach. This is the first report on the characterization of the kefir community associated with a bacteriocin-producing strain. Sequencing data confirmed previous findings using culture dependent approaches that the microbiota of kefir milk and the starter grain are quite different while at the same time, establishing that the microbial diversity of the starter grain is not uniform. An overall greater level of bacterial diversity associated with the interior starter grain compared with the exterior was observed. 16S compositional sequencing also facilitated the identification of inhabitants not previously associated with this complex community. Additionally culture-dependent approaches confirmed the dominance of lacticin 3147-producing lactococci in kefir milk. As the kefir community is a true example of symbiosis, a comprehensive ‘snapshot’ of the bacterial composition, such as that obtained by pyrosequencing-based technology, may begin to aid in the identification and elucidation of the complex interactions associated with this community.


The authors would like to thank Fiona Fouhy for her technical assistance. This work was supported by the Science Foundation of Ireland funded Centre for Science, Engineering and Technology, the Alimentary Pharmabiotic Centre (APC).