Sequence-based analysis of the microbial composition of water kefir from multiple sources


Correspondence: Paul D. Cotter, Teagasc Food Research Centre, Moorepark, Fermoy, Co. Cork, Ireland. Tel.: +353 02542694;

fax: +353 02542340;



Water kefir is a water–sucrose-based beverage, fermented by a symbiosis of bacteria and yeast to produce a final product that is lightly carbonated, acidic and that has a low alcohol percentage. The microorganisms present in water kefir are introduced via water kefir grains, which consist of a polysaccharide matrix in which the microorganisms are embedded. We aimed to provide a comprehensive sequencing-based analysis of the bacterial population of water kefir beverages and grains, while providing an initial insight into the corresponding fungal population. To facilitate this objective, four water kefirs were sourced from the UK, Canada and the United States. Culture-independent, high-throughput, sequencing-based analyses revealed that the bacterial fraction of each water kefir and grain was dominated by Zymomonas, an ethanol-producing bacterium, which has not previously been detected at such a scale. The other genera detected were representatives of the lactic acid bacteria and acetic acid bacteria. Our analysis of the fungal component established that it was comprised of the genera Dekkera, Hanseniaspora, Saccharomyces, Zygosaccharomyces, Torulaspora and Lachancea. This information will assist in the ultimate identification of the microorganisms responsible for the potentially health-promoting attributes of these beverages.


Water kefir is a water–sucrose beverage that is fermented to produce carbon dioxide and low concentrations of ethanol. Traditionally, figs or other dried fruit and lemon are added to provide additional minerals and flavour, with acidity being proportional to the length of the fermentation process. Water kefir is fermented via a combination of bacteria and yeast, which live in symbiosis in water kefir grains. Also known as Tibicos or sugary kefir, water kefir differs from the more popular milk-based kefir in that the grains are principally composed of dextran and are translucent and crystal-like in appearance (Neve & Heller, 2002).

There are two prevalent theories as to the origin of water kefir, but the exact source remains unknown. The first theory is that soldiers returning from the Crimean war brought grains to western Europe (Ward, 1892), while the second proposes the spontaneous formation of grains on the pads of the Opuntia cactus in Mexico (Lutz, 1899). Regardless, water kefir grains were passed from household to household and this is still the most common means by which grains and beverages are acquired as water kefir production has yet to be commercialised on a significant scale.

Water kefir is less studied than other fermented beverages such as milk kefir or kombucha and, as such, any claimed health benefits have yet to be confirmed. Culture-based studies have shown that the bacterial component is comprised of a varied mixture of Lactobacillus, Lactococcus, Leuconostoc and Acetobacter (Pidoux, 1989; Pidoux et al., 1990; Franzetti et al., 1998; Neve & Heller, 2002; Gulitz et al., 2011). Culture-independent methods have also been used to assess the microbial populations of water kefir, including one recent study that used high-throughput sequencing to assess the bacterial component of four water kefir grains (Gulitz et al., 2013). The yeast component has most commonly been identified, via culture-based studies, as Saccharomyces, Hanseniaspora/Kloeckera, Zygotorulaspora and Candida (Pidoux, 1989; Neve & Heller, 2002; Waldherr et al., 2010; Gulitz et al., 2011). It is estimated that bacteria are present in the grains at c. 106–108 per g, with yeast numbers at 106–107 per g (Gulitz et al., 2011).

As a potential functional/health-promoting food, it is noteworthy that water kefir contains strains from species with which health benefits have been frequently attributed, such as lactobacilli, bifidobacteria (Gulitz et al., 2013) and to a lesser extent, Saccharomyces-related yeasts (Czerucka et al., 2007; Foligne et al., 2010). Water kefir therefore represents a potentially effective means of probiotic delivery. With this in mind, the aim of this study was to characterise the bacterial and, to a lesser extent, yeast components within water kefir using high-throughput, culture-independent techniques. This study represents the first occasion upon which high-throughput sequencing has been utilised to investigate the fungal component of water kefir, and together with the corresponding bacterial analysis, provides a comprehensive insight into the microbiological profile of water kefir. It is worth noting that this study reports the microbial composition of water kefir grains composed of dextran, which have not been associated with milk fermentation, unlike a small number of studies reporting water kefir that would appear to have been prepared from milk kefir grains introduced into sucrose–water solution (Miguel et al., 2011; Puerari et al., 2012).

Materials and methods

Culture maintenance

Four water kefir grains were acquired from commercial and individual suppliers from Canada [Ca], the UK [UK] and the United States [US1 and US2]. The cultures were fermented under uniform conditions by adding 60 g grains per L of sterilised Ballygowan® mineral water supplemented with 10% sucrose, followed by the addition of one dried, organic fig (Rainbow Organic Wholefoods, Ireland). The culture was fermented at 25 °C, and after 24 h, DNA was extracted from the fermentate and the grains. Each culture was routinely changed two times per week for a minimum of 10 fermentation cycles prior to extraction and cultures were kept separate from one another to prevent cross-contamination.

Metagenomic DNA extraction

To extract DNA from the fermentate, 1.8 mL was centrifuged to generate a pellet, which was suspended in 450 μL of lysis buffer P1 from the Powerfood Microbial DNA Isolation kit (MoBio Laboratories Inc.). The resuspended pellet was subjected to enzymatic digestion with enzymes mutanolysin (100 U mL−1; Sigma) and lysozyme (50 μg mL−1; Sigma) at 37 °C for 1 h, followed by proteinase K (250 μg mL−1; Sigma) digestion at 55 °C for 1 h. Extraction was optimised with a 10 min 70 °C incubation (Quigley et al., 2012) prior to mechanical lysis. The Powerfood Microbial DNA Isolation kit was then used as per the manufacturer's instructions. For extraction of DNA from the grain, 1 g of grains were washed twice in sterile H2O and homogenised using a mortar and pestle, after which DNA was isolated using a modified phenol–chloroform-based extraction procedure (Garbers et al., 2004).

DNA amplification and pyrosequencing

PCR amplification of the V4-V5 variable region (408 bp) of the 16S rRNA gene was performed using the universal primers F1 (5′-AYTGGGYDTAAAGNG) and R5 reverse (5′-CCGTCAATTYYTTTRAGTTT) to allow an investigation of the bacterial component of the microbial populations (Claesson et al., 2010). Unique multiplex identifier adaptors were attached between the 454 adaptor sequences and the forward primers. Tagged universal primers were also used to amplify fungal DNA from the variable ITS-1 rRNA region (Buee et al., 2009). In this instance, the forward primer ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA) and ITS2 reverse (5′-GCTGCGTTCTTCATCGATGC) generated two different sets of PCR products of circa 410 and 250 bp. To prevent preferential sequencing of the smaller-sized reads, these were pooled and sequenced separately. Amplicons generated from three PCRs of template DNA were pooled and cleaned using the Agencourt AMPure® purification system (Beckman Coulter; Takeley, UK). Sequencing of the 16S rRNA gene V4-V5 and ITS1 rDNA ribosomal amplicons was performed using a 454 Genome Sequencer FLX (Roche Diagnostics Ltd, Burgess Hill, West Sussex, UK) at Teagasc Food Research Centre, Moorepark.

Pyrosequencing data analysis

Raw sequences were quality trimmed and filtered using the Qiime Suite of programmes (Caporaso et al., 2010). Any reads not meeting a minimum quality score of 25 and sequence length shorter than 150 bps for 16S amplicon reads and 200 bps for ITS amplicon reads were discarded. Trimmed fasta sequences were assessed by blast analysis against the silva database (version 100) for 16S reads and a previously published ITS-specific database for ITS samples (Pruesse et al., 2007; Santamaria et al., 2012). blast outputs were parsed using megan (Huson et al., 2007); a bit-score of 86 was employed for 16S ribosomal sequence data, and a bit-score of 35 was used for ITS sequence data. Reads were deposited in the SRA database under the accession number ERP002661.

Results and discussion

High-throughput sequencing reveals the alpha and beta diversity of the bacterial population in water kefir

In this study, both grains and fermentates were analysed, as it has been shown that the grains and fermentates of associated beverages can differ (e.g. milk kefir), with the solid material typically having a more complex microbial population than the final fermented beverage (Marsh et al., 2013). To prevent an origin-based bias, water kefir was sourced from different geographical locations and was fermented to simulate natural artisanal fermenting conditions. DNA was extracted from both the grain and 24-h fermentate to afford the most accurate assessment of total microbial population.

Amplicons corresponding to the V4-V5 variable region of the bacterial 16S rRNA gene were generated using universal tagged primers. To determine species richness and diversity, Chao1 values and Shannon and Simpson indices were calculated, in addition to Phylogenetic Diversity and Observed Species numbers (Supporting Information, Table S1). Values for these were low by comparison to other environmental analyses, highlighting low alpha diversity. Rarefaction curves were calculated at 97% similarity and are approaching parallel to the x-axis for all samples, indicating sufficient reads were obtained to adequately assess the bacterial population (Supporting Information, Fig. S1). Beta diversity was measured using the unweighted UniFrac distance matrix. A principal coordinate analysis (PCoA) plot was generated based on unweighted UniFrac distance matrices to visualise any clustering among the different groups (Fig. S2). No clustering was evident based on sample type or origin, but it was apparent that the data points corresponding to grains of US2 and Ca were each distantly removed from all other data points establishing a quite different beta diversity among these samples. ITS rDNA amplicons were in general sequenced at a lower depth and thus corresponding alpha and beta diversity analyses were not completed.

The bacterial component of water kefir was dominated by the genus Zymomonas, with a LAB and AAB presence

16S reads from both the grain and the fermentate were assigned to three bacterial phyla; Actinobacteria, Firmicutes and Proteobacteria. Across both the grains and fermentates, the dominant phyla were Proteobacteria and Firmicutes. Proportions of Proteobacteria were in general greater in the grain (> 70%) than the fermentates (50-60% abundance for Ca, UK and US1 but 75% abundance for US2; Table S2). Proportions of Firmicutes were consistently higher in the fermentates than in the grains. The abundance of Actinobacteria was low in general but was higher in each of the grains than in the fermentates.

The most striking difference between a recent high-throughput sequencing-based study (Gulitz et al., 2013) and data presented here relates to the fact that in this study the most abundant genus was Zymomonas. Indeed, Zymomonas was dominant in all grains and fermentates, with proportions being consistently higher in the grains. Proportions of Zymomonas ranged from 87% in the grain of US1 to 49.5% in the fermentate (Fig. 1). Zymomonas is traditionally regarded as a fermentation-associated microorganism and forms part of the microbial population of several fermented beverages from plants in tropical regions of America, Africa and Asia. It has been isolated from Mexican pulque, a fermented beverage made from agave sap which resembles water kefir if theories with respect to the Opuntia cactus are correct (Ward, 1892). Additionally, it has already been established that Zymomonas can constitute part of the microbial population of water kefir (Hsieh et al., 2012), albeit not at such high levels.

Figure 1.

16S phylogenetic composition of the bacterial component of the water kefir grain (a) and kefir fermentate (b) at genus level.

Zymomonas is capable of producing ethanol in such high quantities that it rivals the fermentation capabilities of Saccharomyces (Panesar et al., 2006). Zymomonas species have also been shown to produce levan, a polysaccharide used as a thickening agent, which has been demonstrated to have antitumor effects (Calazans et al., 1997; Yoon et al., 2004), immunostimulating activity (Xu et al., 2006), effects on lipid metabolism (Yamamoto et al., 1999) and prebiotic activity (Dal Bello et al., 2001; Jang et al., 2003). Zymomonas itself has been shown to be safe for consumption (de Azeredo et al., 2010) and to have beneficial immunomodulatory effects by protecting against sepsis in mice (Campos et al., 2013). Further analysis is required to characterise the strains of Zymomonas present in these beverages and their contribution, through the production of leaven, sorbitol and ethanol, on the consistency, flavour and safety of the final beverage. Nonetheless, at the significant proportions detected, members of this genus are undoubtedly of considerable importance with respect to the composition of these beverages. It should be noted that while Zymomonas represents a dominant component of these water kefirs, additional studies with a larger number of samples will be required to assess the distribution of this genus in beverages from other regions and prepared in different ways.

The majority of non-Zymomonas reads from the water kefir grains and fermentates were assigned to the lactic acid bacteria (LAB) and Lactobacillus in particular. This genus has long been associated with water kefir and in a number of instances has been reported to be the dominant bacterial genus present (Gulitz et al., 2011, 2013). Here, Lactobacillus accounted for 25.4%, 16.4%, 12% and 22.3% of the grains of Ca, UK, US1 and US2, respectively, and 38.8%, 38.5%, 23.4% and 25.1% of the corresponding fermentates, respectively, and thus proportions increased from the grain to the fermentate in all instances. Lactobacillus hilgardii was believed to be the primary exopolysaccharide producer in water kefir (Pidoux, 1989; Waldherr et al., 2010), whereas Gulitz et al. reported L. nagelii and L. hordei to be the major EPS producers (Gulitz et al., 2011). Leuconostoc was recognised as the major LAB in the fermentate of US1 (25.6%), but was less abundant in the grain environment (Ca, 1.1%; US1, 0.4%). Thus, while water kefir-associated Leuconostoc has been shown to be competitive and to be capable of EPS-production (Gulitz et al., 2011), its detection in such low grain abundance suggests that this genus does not contribute significantly to grain formation in these samples. No lactococci were detected in our study, despite representatives of this genus having been isolated as part of culture-based studies of other water kefirs previously (Pidoux, 1989; Waldherr et al., 2010).

The genera Acetobacter and Gluconacetobacter were represented as a minor proportion of the overall bacterial component of water kefir (Fig. 1). Gluconacetobacter were present in the grains of Ca, US1 and US2 at < 1% and were undetected in the fermentates and UK grain, whereas, interestingly, Acetobacter appeared to be better adapted to the fermentate (Ca, 0.4%; UK, 2.8%; US1, 1.1% and US2, 0.6%). Acetic acid bacteria have been consistently detected at trace levels in water kefir studies (Franzetti et al., 1998; Gulitz et al., 2011, 2013) but, as yet, no role has been attributed to them.

Members of the Bifidobacteriaceae family were detected at low proportions, but could not be confidently assigned to genus level. The presence of Bifidobacterium in water kefir was only recently revealed via high-throughput sequencing (Gulitz et al., 2013), but had not been detected by previous culture- or molecular-based analyses. Our analysis showed levels of Bifidobacteriaceae were lower than in the fermentate (< 2%) than the grain (Ca, 3.6%; UK, 7.8%; US1, 0.2% and US2, 2.5%), perhaps indicative of poor survival of representatives from this family in a non-EPS protected environment.

Insight into the fungal composition of water kefir and associated water kefir grain

As the primary focus of this study was the investigation of the bacterial composition of kefir grains and beverages, fungal (ITS) amplicons were sequenced at a much lower depth with a view to providing an initial insight into this population. Amplification with ITS primers yielded two amplicons of c. 250 and 410 bp, respectively, which were purified and sequenced separately. The existence of two pools limited, to a certain extent, interpool comparisons but the dominant components of the fungal population were nonetheless evident. All reads were assigned to the phylum Ascomycota. At the family level, Saccharomycetaceae and Saccharomycodaceae were detected in both the grain and water kefir fermentate, while Debaromycetaceae were detected in the fermentate only (Table S3). Almost all reads from the 250 bp pool were assigned to the genus Dekkera while the 410 bp pool contained amplicons corresponding to a variety of species.

Dekkera was represented by two species, Dekkera anomala and Dekkera bruxellensis. Both are common fermentation-associated species, with D. bruxellensis frequently associated with the spoilage of fermented beverages (Heresztyn, 1986). To our knowledge, Dekkera has not been previously identified in water kefir but has been found in other fermented beverages (Pintado et al., 1996; Wyder et al., 1997; Miyamoto et al., 2010). The fact that Dekkera has previously been overlooked by culture-based approaches may in part be due to the fact that it has a slow doubling time when cultured on standard yeast media and, thus, can be outcompeted by other genera such as Saccharomyces (Ibeas et al., 1996).

From the 410 bp pool, proportions of Saccharomyces were highest in the fermentate of US1. Saccharomyces were also detected in the fermentates of Ca and US2, despite having not been detected in the corresponding grains (Table S3). Saccharomyces could only be assigned to the species level among ITS reads corresponding to the UK grain, with 1.3% of reads being assigned as Saccharomyces cerevisiae. This has previously been identified as a key yeast in several water kefirs (Franzetti et al., 1998; Gulitz et al., 2011). Despite this, the proportions of Saccharomyces detected in our study were lower than expected. Waldherr et al. also noted this phenomenon, which suggests that levels of Saccharomyces may have been overestimated in previous studies (Waldherr et al., 2010). The genus Hanseniaspora, whose anamorphic form is Kloeckera, has frequently been detected in water kefir, most often as Hanseniaspora valbyensis and Hanseniaspora vineae (Galli et al., 1995; Franzetti et al., 1998; Waldherr et al., 2010; Gulitz et al., 2011). Both species were also detected in our study. While H. vineae was detected in more samples, it was generally (with the exception of US1) present at a lower abundance than H. valbyensis. H. valbyensis was most prominent in the Ca fermentate and was also present in the US1 fermentate and UK grain. The genus Lachancea was not detected in the fermentate, but was found in the grain of UK, with all reads being assigned to Lachancea fermentati. This species was formerly known as Zygosaccharomyces fermentati (Kurtzman, 2003) and has been detected in water kefir previously (Gulitz et al., 2011). Torulaspora was detected in the grain of UK and the fermentates of US1 and US2, but could not be assigned at the species level. Torulaspora has been identified in water kefir on only one previous occasion as Torulaspora pretoriensis (Pidoux, 1989). Zygotorulaspora florentina [formerly Zygosaccharomyces florentinus/Saccharomyces florentinus (Kurtzman, 2003)] was unexpectedly not identified in any of the cultures, despite frequently appearing as a dominant species in previous analyses (Pidoux, 1989; Neve & Heller, 2002; Gulitz et al., 2011). While it is possible that it was present in lower abundance and thus was below our detection limits, this species probably does not exert the positive impact on the growth of lactobacilli that has been attributed to it in previous studies (Leroi & Pidoux, 1993a,b; Stadie et al., 2013). It may be that another species of yeast performs this role within the samples assessed on this occasion.

Finally, two other yeast have been detected, which have not previously been associated with water kefir. One species of Zygosaccharomyces, Zygosaccharomyces lentus, was detected at low levels in the fermentate of UK and grains of Ca, US1 and US2. Z. lentus is often considered a spoilage yeast (Steels et al., 1999), and this is the first report of the detection of this species in water kefir. Meyerozyma was detected in the fermentates of US1 and US2, but of these reads, only 0.3% from US2 could be assigned at the species level (from the 250 bp pool; Table S4) to Meyerozyma caribbica.

In conclusion, we have established that the bacterial population of a number of water kefirs analysed through high-throughput sequencing consists of a dominant population of Zymomonas with lactic acid bacteria (Lactobacillus, Leuconostoc), acetic acid bacteria (Acetobacter and Gluconacetobacter) and Bifidobacteriaceae being detected in descending proportions. While dextran-producing lactic acid bacteria are not the dominant bacteria, it would seem that the possibility exists for the modulation of the microbiology of water kefir through the introduction of nonindigenous lactobacilli and bifidobacteria to facilitate its use as a nondairy-based system for the delivery of probiotics. Our results revealed that the yeast component of the water kefir samples was comprised of several species previously associated with water kefir, but notably a number of species not traditionally associated with water kefir, including Dekkera, Zygosaccharomyces and Meyerozyma were also identified.


The Alimentary Pharmabiotic Centre is a research centre funded by Science Foundation Ireland (SFI), through the Irish Government's National Development Plan. The authors and their work were supported by SFI CSET grant APC CSET 2 grant 07/CE/B1368. The authors would like to thank Eva Rosberg-Cody and Fiona Crispie for technical assistance.