Journal of Applied Microbiology ISSN 1364-5072

ORIGINAL ARTICLE

The water kefir grain inoculum determines the

characteristics of the resulting water kefir fermentation
process
D. Laureys and L. De Vuyst
Research Group of Industrial Microbiology and Food Biotechnology, Faculty of Sciences and Bioengineering Sciences, Vrije Universiteit Brussel,
Brussels, Belgium

Keywords
aroma, bifidobacteria, exopolysaccharide,
lactic acid bacteria, water kefir, yeasts.
Correspondence
Luc De Vuyst, Research Group of Industrial
Microbiology and Food Biotechnology, Faculty
of Sciences and Bioengineering Sciences, Vrije
Universiteit Brussel, Pleinlaan 2, B-1050 Brus-
sels, Belgium.
E-mail: ldvuyst@vub.ac.be
2016/1747: received 11 August 2016, revised
20 November 2016 and accepted 29 Novem-
ber 2016
doi:10.1111/jam.13370
Abstract
Aims: To investigate the influence of the water kefir grain inoculum on the
characteristics of the water kefir fermentation process.
Methods and Results: Three water kefir fermentation processes were started
with different water kefir grain inocula and followed as a function of time
regarding microbial species diversity, community dynamics, substrate
consumption profile and metabolite production course. The inoculum
determined the water kefir grain growth, the viable counts on the grains, the
time until total carbohydrate exhaustion, the final metabolite concentrations
and the microbial species diversity. There were always 2–10 lactic acid bacterial
cells for every yeast cell and the majority of these micro-organisms was always
present on the grains. Lactobacillus paracasei, Lactobacillus hilgardii,
Lactobacillus nagelii and Saccharomyces cerevisiae were always present and may
be the key micro-organisms during water kefir fermentation. Low water kefir
grain growth was associated with small grains with high viable counts of
micro-organisms, fast fermentation and low pH values, and was not caused by
the absence of exopolysaccharide-producing lactic acid bacteria.
Conclusions: The water kefir grain inoculum influences the microbial species
diversity and characteristics of the fermentation process. A select group of key
micro-organisms was always present during fermentation.
Significance and Impact of the Study: This study allows a rational selection of
a water kefir grain inoculum.

organisms (Moinas et al. 1980; Laureys and De Vuyst

Introduction
Water kefir is a fermented beverage that is made by
adding water kefir grains (the inoculum) to a mixture
of water, (dried) fruits and sugar (Gulitz et al. 2011;
Marsh et al. 2013; Laureys and De Vuyst 2014). This
mixture is fermented for 2–4 days at room temperature
under anaerobic conditions, followed by sieving to sep-
arate the grains, which are reused for the next fermen-
tation process through a backslopping procedure, from
the liquor. The liquor is slightly sweet, acidic, alcoholic
and sparkling, with a yellowish colour and a fruity
taste and aroma. The grains are brittle, consist of dex-
tran exopolysaccharides (EPS) and harbor the micro-
2014).
The main micro-organisms found in water kefir are
lactic acid bacteria (LAB), yeasts, bifidobacteria and acetic
acid bacteria (Magalh~aes et al. 2010, 2011; Gulitz et al.
2011, 2013; Laureys and De Vuyst 2014). Different water
kefirs harbour different species diversities, but it is still
unclear which are the key micro-organisms during a fer-
mentation course and how species diversity influences the
fermentation process. The LAB species Lactobacillus hil-
gardii is frequently associated with water kefir fermenta-
tion and is thought to be responsible for the water kefir
grain growth because of its production of EPS from
sucrose (Pidoux 1989; Waldherr et al. 2010). Recently, a

Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology 719
Impact of the water kefir inoculum
novel Bifidobacterium species was found in water kefir,
but its importance during water kefir fermentation
remains unclear (Laureys et al. 2016). The main metabo-
lites produced during water kefir fermentation are etha-
nol, lactic acid, glycerol, acetic acid and mannitol; the
main aroma compounds are 2-methyl-1-propanol,
isoamylalcohol, ethyl acetate, isoamyl acetate, ethyl hex-
anoate and ethyl octanoate (Laureys and De Vuyst 2014).
Currently, the water kefir beverage is predominantly
made at a household level, whereby the grains are handed
over from person to person (Laureys and De Vuyst
2014). This practice is possible because the water kefir
grain mass normally increases upon fermentation. From a
commercial point of view, the water kefir fermentation
process is difficult to control, as it can become unstable,
yielding variable end-products. Also, the water kefir grain
growth often decreases, which prevents successful back-
slopping or upscaling of the production process. To be
able to avoid and/or remedy these problems during fer-
mentation and to allow the development of a stable com-
mercial production process, a thorough understanding of
the water kefir fermentation process is required.
This comparative study aimed to elucidate the influ-
ence of the water kefir grain inoculum on microbial spe-
cies diversity, community dynamics, pH evolution, water
kefir grain growth, substrate consumption profile and
metabolite production course during fermentation, and
to make potential links between certain fermentation
characteristics.
Materials and methods
Water kefir grain inocula and prefermentations
Three water kefir grain inocula (A, B and C) were
obtained from different private persons that carry out
water kefir fermentations at home (Table 1). Each of
Table 1 Approximate production characteristics of the water kefir fermentation processes carried out by private persons at household level, from
which the grain inocula for the processes A, B and C were obtained for the present comparative study of the influence of the water kefir grain
inoculum. The concentrations are given per litre of water used in the recipe
Water kefir grain inoculum
Characteristic A B
Origin Leuven, Belgium
Water kefir grains (g l 1) 90
Sugar (g l 1) 110
Fruits and other ingredients (l 1) Two dried figs
Three dried apricots
2 ml of apple cider vinegar
Fermentation conditions 15°C, 2 days
Estimated grain growth (%) 20
720 Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology
D. Laureys and L. De Vuyst
these inocula (100 g) was cultivated through a series of
consecutive prefermentations (through backslopping) in
Schott bottles until >600 g of water kefir grain wet mass
was obtained (Laureys and De Vuyst 2014). The Schott
bottles (1, 2 and 5 l) were equipped with a polytetrafluo-
roethylene water lock and incubated in a water bath at
21°C. Per 15 g of water kefir grains, 85 ml of autoclaved
(121°C, 2
1 bar, 21 min) water kefir simulation medium
(WKSM) was added. The WKSM contained 6 g of unre-
fined cane sugar (Candico Bio, Merksem, Belgium),
65 ml of distilled water and 20 ml of fig extract, which
was prepared as described previously (Laureys and De
Vuyst 2014). After 72 h, the backslopping practice was
applied, whereby the grains were separated from the
liquor by sieving and recultivated in fresh WKSM under
the same conditions as described above.
Fermentations
The inocula obtained after a series of prefermentations
were used to start the water kefir fermentation processes
in triplicate. Therefore, 100-ml Schott bottles (equipped
as described above and several bottles per fermentation)
were filled with 85 ml of autoclaved WKSM and 15 g of
water kefir grain inoculum. Incubation of the bottles was
performed in a water bath at 21°C. After closure of the
bottles at the start of the fermentations as well as before
their sampling, the contents were mixed by mildly turn-
ing them. For water kefirs B and C, three bottles (repre-
senting three independent biological replicates) were
removed and their contents analysed after 0, 6, 12, 18,
24, 48, 72, 96, 144 and 192 h of fermentation. For water
kefir A, only one bottle was removed and its contents
analysed in triplicate (representing three technical repli-
cates) after 0, 24 and 72 h of fermentation, and three
bottles (representing three independent biological repli-
cates) were removed and their contents analysed after
C
Schiedam, the Netherlands Lokeren, Belgium
100 250
90 60
One dried fig Two dried figs
20 g of raisins
One slice of peeled lemon
19°C, 3 days 20°C, 3 days
25 50
D. Laureys and L. De Vuyst
48 h of fermentation, because the water kefir grain wet
mass of inoculum A did not increase during the series of
prefermentations. The analysis results at each sampling
point are presented as the mean  standard deviation of
the three measurements.
Water kefir grain wet and dry mass and pH
determinations
The water kefir grain wet mass was measured after sieving
the contents of a bottle and washing the grains with
200 ml of sterile saline (8
5 g l 1 of NaCl (Merck, Darm-
stadt, Germany)). The water kefir grain growth was
defined as the increase in the grain wet mass at the time
of sampling (compared with that at the start of the fer-
mentation) divided by the grain wet mass at the start of
the fermentation, and expressed as percentage (%, m/m).
The water kefir grain dry mass was measured after drying
of 5 g of washed grain wet mass at 105°C for 48 h, and
defined as the mass after drying divided by the mass
before drying, expressed as % (m/m). The pH of the
water kefir liquor was measured with a SenTix 41 glass
electrode (WTW, Weilheim, Germany), immediately after
a bottle was opened.
Microbial enumerations
To enumerate micro-organisms in the liquors and on the
grains, decimal dilutions of samples of liquor and grain
suspension were prepared and plated on different agar
media, as described previously (Laureys and De Vuyst
2014). Presumptive LAB and yeasts were enumerated on
de Man–Rogosa–Sharpe (MRS) agar medium supple-
mented with cycloheximide (final concentration of
0
1 g l 1; Sigma-Aldrich, St Louis, MO) and yeast
extract-glucose (YG) agar medium supplemented with
chloramphenicol (final concentration of 0
1 g l 1; Sigma-
Aldrich), respectively, after incubation at 30°C for 4 days.
After 0 and 72 h of fermentation, the samples were also
plated on kanamycin aesculin azide (KAA) and violet red
bile glucose (VRBG) agar media to enumerate presump-
tive enterococci plus streptococci and Enterobacteriaceae,
respectively, after incubation at 42°C for 24 h. Colony
enumerations were expressed as log CFU per millilitre of
liquor or per gram of grains.
Culture-dependent species diversity and community
dynamics analyses
The culture-dependent species diversity and community
dynamics analyses were performed after 0 and 48 h of
fermentation for water kefir A, and after 0, 48 and
192 h of fermentation for water kefirs B and C. Hereto,
Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology
Impact of the water kefir inoculum
10–20% of the colonies on MRS and YG agar media with
a total colony count between 30 and 300 were randomly
picked up and purified on their respective agar media.
Purified isolates were cultured in the appropriate liquid
media to obtain cell pellets. These were treated with
mutanolysin, lysozyme and proteinase K in the case of
bacteria, and lyticase and proteinase K in the case of
yeasts to extract DNA, as described previously (Laureys
and De Vuyst 2014). The DNA was purified with a
Nucleospin 96 tissue kit (Macherey-Nagel, D€ uren, Ger-
many), according to the instructions of the manufacturer.
Purified bacterial and yeast DNA was diluted until
50 ng ll 1 and subjected to (GTG)5-PCR- and M13-
PCR-fingerprinting respectively (Laureys and De Vuyst
2014). The fingerprint patterns obtained were clustered
numerically with Bionumerics ver. 5 software (Applied
Maths, Sint-Martens-Latem, Belgium). Representative iso-
lates within each cluster were identified by amplifying
and sequencing part of the 16S rRNA gene (primer pair
pA/pH) for bacteria (Edwards et al. 1989), and part of
the large subunit of the 26S rRNA gene (primer pair
LR0R/LR3) (Vilgalys and Hester 1990) and the internal
transcribed spacer (ITS) region (primer pair ITS1/ITS4)
(White et al. 1990) for yeasts. The closest known type
strains of the sequenced fragments were determined
(expressed as % identity) with the BLAST algorithm
(Altschul et al. 1990) and the GenBank database (http://
blast.ncbi.nlm.nih.gov/). The accession numbers of their
gene sequences are reported.
All bacterial isolates were grown on MRS agar medium
supplemented with 10 g l 1 of sucrose at 30°C for 7 days
to visually assess their EPS production capacity.
Culture-independent species diversity and community
dynamics analyses
The culture-independent species diversity and community
dynamics analyses were performed after 0, 24, 48 and
72 h of fermentation for water kefir A, and after 0, 24,
48, 72 and 192 h of fermentation for water kefirs B and
C. Hereto, 40 ml of liquor and 10 ml of grain suspension
were centrifuged (7200 g, 20 min, 4°C) and the super-
natants were discarded. DNA from the pellets was
obtained after treatment with lyticase, mutanolysin, lyso-
zyme and proteinase K, as described previously (Laureys
and De Vuyst 2014). The DNA extract was further puri-
fied with a Nucleospin food kit (Macherey-Nagel),
according to the instructions of the manufacturer. Puri-
fied DNA was diluted until 50 ng ll 1 and used for
PCR-DGGE with primer pairs 357f-GC/518r for bacteria
(V3), LAC1/LAC2-GC for LAB (LAC), bif164f/bif662r-
GC for bifidobacteria (Bif) and NL1-GC/LS2 for yeasts
(Yeast); the indication GC refers to a GC clamp. The
721
Impact of the water kefir inoculum
PCR amplicons were separated in a 6% (v/v) polyacry-
lamide gel with a denaturing gradient from top to
bottom of 45–60% (V3 and Yeast), 40–55% (LAC) and
45–55% (Bif), as described previously (Papalexandratou
et al. 2011). Selected bands were cut from the gel, ampli-
fied with their respective primer pairs (without GC
clamp) and sequenced for identification. The closest rela-
tives of the sequenced fragments (expressed as % iden-
tity) were identified as described above.
Substrate consumption and metabolite production
profile analyses
Cell-free supernatants were obtained after centrifugation
(7200 g, 20 min, 4°C) of the sieved liquors. They were
pretreated (except for the analysis of aroma compounds)
by vortexing, centrifugation (21 000 g, 20 min, 4°C) and
filtration (0
2-lm pore size Whatman filters; GE Health-
care Life Sciences, Bucks, UK) before injection into a col-
umn. Quantifications were performed with external
calibration curves with standards prepared in the same
way as the samples.
Concentrations of sucrose, fructose and glucose were
measured through high-performance anion exchange
chromatography with pulsed amperometric detection
(HPAEC-PAD) equipped with a CarbopacTM PA10 col-
umn (Dionex, Sunnyvale, CA), as described previously
(Laureys and De Vuyst 2014). These samples were treated
as described above, injected (10 ll) into the column and
eluted as described previously (Janssens et al. 2012). Con-
centrations of acetic acid and D- and L-lactic acid were
measured through high-performance liquid chromatogra-
phy with ultraviolet detection (HPLC-UV) with a Waters
chromatograph (Waters, Milford, MA) equipped with a
Shodex ORpak CRX-853 column (Showa Denko, Tokyo,
Japan) coupled with a UV detector operating at 253 nm
(Waters). Therefore, 250 ll of cell-free supernatant was
added to a mixture of 500 ll of acetonitrile and 250 ll
of ultrapure water. These samples were treated as
described above, injected (30 ll) into the column and
eluted with 10% acetonitrile and 90% 1 mmol l 1
CuSO4. Concentrations of glycerol and mannitol were
measured through HPAEC-PAD with the same Dionex
chromatograph as mentioned above but equipped with a
Carbopac MA1 column (Dionex), as described previously
(Laureys and De Vuyst 2014). These samples were treated
as described above, injected (10 ll) into the column and
eluted as described previously (Wouters et al. 2013b).
Concentrations of ethanol were measured through gas
chromatography with flame ionization detection with a
Focus gas chromatograph (Interscience, Breda, the
Netherlands) equipped with a Stabilwax-DA column
(Restek, Bellefonte, PA), as described previously (Laureys
722 Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology
D. Laureys and L. De Vuyst
and De Vuyst 2014). Therefore, 100 ll of cell-free super-
natant was added to 1100 ll of sample preparation solu-
tion (600
0 ll of acetonitrile, 367
8 ll of ultrapure water,
12
0 ll of formate and 0
2 ll of 4-methyl-2-pentanol
(internal standard; Sigma-Aldrich)). These samples were
treated as described above, injected (1 ll) into the col-
umn and eluted as described previously (Rimaux et al.
2011).
Volatile aroma compounds in the headspace of the
water kefir liquor were determined through static head-
space (SH) and solid phase micro-extraction (SPME) gas
chromatography with mass spectrometry detection (GC-
MS) by means of an Agilent 6890 chromatograph
equipped with a DB-WAXetr column coupled with an
Agilent 5973N mass spectrometer (Agilent Technologies,
Santa Clara, CA). Sample preparation was as described
previously (Laureys and De Vuyst 2014). For the mea-
surements via SH-GC-MS, the samples were equilibrated
at 40°C for 30 min at 400 rev min 1 in a MPS2
autosampler (Gerstel, M€ ulheim an der Ruhr, Germany)
before 1
0 ml of headspace was injected into the column
and eluted as described previously (Wouters et al.
2013a). For the measurements via SPME-GC-MS, a
divinylbenzene/carboxen/polydimethylsiloxane fibre
(Supelco, Bellefonte, PA) was equilibrated in the water
kefir liquor headspace at 40°C for 30 min at
400 rev min 1 in the MPS2 autosampler before desorp-
tion of the compounds from the fibre into the column
and elution as described previously (Leroy et al. 2009).
The compounds were identified by comparison of the
mass spectra with library data (NIST 08 database: http://
www.nist.gov) and comparison of the retention times of
reference compounds (if available). Relative abundances
were calculated by normalization of the peak areas with
the internal standard and multiplication with a factor of
1000.
Composition analysis of the water kefir grain
exopolysaccharides
The composition of the water kefir grain EPS was anal-
ysed by suspending 5
0 g of rinsed water kefir grain wet
mass in 45 ml of ultrapure water, shaking by inversion
for 5 min and centrifuging (7200 g for 15 min), after
which the supernatant was discarded. For acid hydrolysis,
0
50  0
01 g of water kefir grain wet mass was added to
5 ml of 2
0 mol l 1 HCl or 2
0 mol l 1 H2SO4 (Emaga
et al. 2012). These mixtures were incubated at 100°C for
1, 2, 4 and 6 h. Carbohydrates in the resulting solutions
were measured via HPAEC-PAD, as described above, but
without adding the internal standard. Organic acids in
the resulting solutions were measured through HPLC-
UV, as described above, and through HPAEC with the
D. Laureys and L. De Vuyst
same Dionex chromatograph as mentioned above but
equipped with a IonPac AS19 column and coupled with
a conductivity under ion suppression detector (Dionex),
as described previously (Moens et al. 2014).
Carbon recovery
Calculations of the carbon recoveries were based on the
measurements of the water kefir grain wet and dry mass
and sucrose, glucose, fructose, ethanol, lactic acid, glyc-
erol, acetic acid and mannitol concentrations, as
described previously (Laureys and De Vuyst 2014).
Statistics
An ANOVA was performed to test for differences between
the fermentation processes, followed by a series of post
hoc pairwise comparisons with Fisher’s least significant
difference test (de Winter 2013). Two-tailed Spearman
correlation coefficients between test variables were calcu-
lated. All statistical tests were performed in R 3
2
0 with
a significance level of 0
05.
Results
Water kefir grain wet mass and pH
The water kefir grain growth of water kefir A was very
low at the end of all prefermentations (Fig. 1). Conse-
quently, only a small amount of water kefir grain wet
mass was available to start the subsequent fermentations.
The water kefir grain growth of water kefir B was low at
the end of the first prefermentation, increased until
100 Water kefir grain growth (%)
80
60
40
20
0
1 2 3 4 5 6 7 8
Backslopping
Figure 1 The water kefir grain growth after 72 h of fermentation
(%) during each of the prefermentations performed to produce the
water kefir grain inocula A (■), B ( ) and C (□); and after 72 h of
fermentation during the water kefir fermentation processes inocu-
lated with the water kefir grain inocula A (●), B ( ) and C (○).
Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology
Impact of the water kefir inoculum
prefermentation 5, and decreased afterwards. The water
kefir grain growth of water kefir C was high at the end of
the first prefermentation and gradually decreased after-
wards. The water kefir grain growth during the actual
water kefir fermentation processes was in line with that
during their prefermentations (Fig. 1).
The water kefir grain growth during water kefir fer-
mentation process A remained very low compared to that
of processes B and C. The water kefir grain growth dur-
ing processes B and C remained comparable during the
first 48 h of fermentation, after which it stopped in water
kefir B and continued in water kefir C until 144 h
(Fig. 2). The water kefir grain growth always stopped
when sucrose was depleted. When the total residual car-
bohydrate (TRC) concentrations were <1 g l 1, it was
significantly different between the three fermentation pro-
cesses (Table 2). The water kefir grain dry mass initially
increased during all three fermentation processes, after
which it decreased along with the decrease in the TRC
concentrations to reach a stable value when the latter
were <1 g l 1. Although the water kefir grain dry mass of
the grains was significantly lower in water kefir A than in
water kefirs B and C, the differences were small
(Table 2). The grains of water kefir A were noticeably
smaller and less transparent than those of water kefirs B
and C.
The pH of the WKSM (before inoculation) was
4
82  0
02. The pH of the liquor decreased fastest in
water kefir fermentation process A and slowest in process
B (Fig. 2). After this fast initial decrease, the pH contin-
ued to decrease slowly along with the increase in the lac-
tic acid and acetic acid concentrations until the end of
the fermentation processes.
Microbial enumerations
Immediately after the water kefir grain inoculum was
added to the WKSM and the bottles were turned mildly,
the viable counts on the MRS and YG agar media for
both liquors and grains plateaued at a certain level and
remained stable during the entire fermentation processes.
The values shown in Table 3 when the TRC concentra-
tions were <1 g l 1 were representative for these entire
processes. The viable counts of the LAB and yeasts on the
grains were significantly different between the three fer-
mentation processes, being highest on these of water kefir
9 A. This was also reflected in the viable counts in the
liquors, albeit less pronounced. When the TRC concen-
trations were <1 g l 1, the water kefir grain growth was
negatively correlated with the viable counts of the LAB
( 0
828; P = 0
006) and yeasts ( 0
895; P = 0
011) on
the grains, but not with those in the liquors. The viable
counts of the LAB on the grains correlated positively with
723
Impact of the water kefir inoculum D. Laureys and L. De Vuyst

50 Sucrose (g l–1) 4·5 pH

40
4·0
30

20
3·5
10

0 3·0
0 48 96 144 192 0 48 96 144 192
20 Glucose (g l–1) 3·0 Lactic acid (g l–1)
2·5
15
2·0

10 1·5

1·0
5
0·5

0 0·0
0 48 96 144 192 0 48 96 144 192

20 Fructose (g l–1) 1·2 Acetic acid (g l–1)

1·0
15
0·8

10 0·6

0·4
5
0·2

0 0·0
0 48 96 144 192 0 48 96 144 192
30 Water kefir grain mass (g) 0·3 Mannitol (g l–1)

25 0·2

20 0·1

15 0·0
0 48 96 144 192 0 48 96 144 192

40 Ethanol (g l–1) 2·5 Glycerol (g l–1)

2·0
30
1·5
20
1·0
10
0·5

0 0·0
0 48 96 144 192 0 48 96 144 192
Time (h) Time (h)

Figure 2 The concentrations of substrates and metabolites, the water kefir grain wet mass and the pH as a function of time during the water
kefir fermentation processes inoculated with water kefir grain inocula A (●), B ( ) and C (○).

724 Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology
D. Laureys and L. De Vuyst Impact of the water kefir inoculum

Table 2 Characteristics of the water kefir fermentation processes inoculated with water kefir grain inocula A, B and C. Statistically significant dif-
ferences (P < 0
05) are indicated with superscript letters

Water kefir fermentation process

Characteristic A B C

Time when (sucrose) <1 g l 1 (h) 24 48 144

Time when (total carbohydrates) <1 g l 1 (h) 48 72 144
Water kefir grain growth (%) 4
58  1
57a 43
98  1
85b 63
82  1
19c
Water kefir grain dry mass (%) 12
87  0
16a 14
41  0
11b 14
52  0
02b
pH 3
34  0
03a 3
47  0
01b 3
35  0
01a
Ethanol (g l 1) 33
77  3
26a 34
10  0
97a 27
04  2
69b
Lactic acid (g l 1) 2
36  0
03a 1
93  0
07b 2
36  0
19a
Acetic acid (g l 1) 0
43  0
04a 0
25  0
03a 0
90  0
16b
Glycerol (g l 1) 2
02  0
18a 1
95  0
04a 2
12  0
02a
Mannitol (g l 1) 0
16  0
02a 0
12  0
01b 0
24  0
01c
Ratio glycerol/ethanol (mmol mol 1) 30  1a 29  1a 39  4b
Ratio lactic acid/ethanol (mmol mol 1) 36  4a 29  1b 45  4c
Ratio acetic acid/ethanol (mmol mol 1) 9
9  1
8a 5
6  0
5a 25
6  4
6b
Ratio acetic acid/lactic acid (mol mol 1) 0
27  0
02a 0
19  0
01b 0
57  0
06c
D-lactic acid (% of total lactic acid) 36
75  0
82a 39
65  1
59b 39
98  0
61b
Carbon recovery (%) 103
9  7
2a 106
7  2
3a 98
1  5
3a

Table 3 Viable counts of the yeasts and lactic acid bacteria in the liquors (log CFU per ml) and on the grains (log CFU per g) of the water kefir
fermentation processes inoculated with water kefir grain inocula A, B and C, when the total carbohydrate concentrations were <1 g l 1. The
ratios between the different viable counts were calculated. Therefore, total yeasts (CFU) and total bacteria (CFU) were calculated based on their
viable counts in the liquors and on the grains taking into account the water kefir grain wet mass (g) and volume of the water kefir liquor (ml).
Statistically significant differences (P < 0
05) are indicated with superscript letters

Water kefir fermentation process

Viable counts or ratio A B C

Yeasts (liquor) 6
44  0
08a

6
72  0
16 b
6
11  0
14a
Bacteria (liquor) 6
92  0
05a 6
86  0
14a,b 6
68  0
15b
Yeasts (grains) 8
26  0
02a 8
03  0
10b 7
68  0
09c
Bacteria (grains) 8
84  0
07a 8
57  0
07b 8
22  0
07c
Bacteria/yeasts (liquor) 3
01  0
42a,b 1
56  0
86a 4
02  1
73b
Bacteria/yeasts (grains) 3
78  0
49a 3
47  0
75a 3
58  1
26a
Grains/liquor (yeasts) 66
08  22
84a 22
84  11
36b 39
57  15
91b
Grains/liquor (bacteria) 82
58  9
78a 53
10  15
49b 35
93  10
90b
Grains/liquor (total yeasts) 12
32  2
09a 6
28  3
07a 12
87  5
10a
Grains/liquor (total bacteria) 15
39  2
08a 14
66  4
39a 11
69  3
47a

those of the yeasts (0
933; P < 0
001), but this was not
the case in the liquors (0
427; P = 0
252). No colonies
were found on the KAA and VRBG agar media, indicat-
ing the absence of enterococci plus streptococci and of
Enterobacteriaceae respectively.
The ratios of the viable counts of the LAB to those of
the yeasts in the liquors and on the grains remained
between 2 and 10 during the entire course of the three
fermentation processes. The ratios of the viable counts of
the LAB and the yeasts on the grains (CFU per g) to
those in the liquors (CFU per ml) remained also more or
less stable between 10 and 100 during the entire fermen-
tation processes. When the grain wet masses and the
liquor volumes were taken into account, the ratios of the
total numbers (expressed as total CFU) of the LAB and
yeasts on the grains to those in the liquors varied
between 5 and 20 during the entire fermentation pro-
cesses. The ratios of the viable counts when the TRC con-
centrations were <1 g l 1 were representative for the
entire fermentation processes A, B and C.
Culture-dependent species diversities and community
dynamics
The species diversities in the liquors and on the grains
were similar and stable as a function of time during the

Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology 725
Impact of the water kefir inoculum D. Laureys and L. De Vuyst

entire course of the three fermentation processes. The

species diversities at the different sampling points for
each water kefir were pooled to display the species diver-
sities of the fermentation processes (Fig. 3). Lactobacillus
paracasei was found in the liquors and on the grains of
water kefirs A, B and C, with similar relative abundances
in the liquors and on the grains. Lactobacillus hilgardii
was found in water kefirs A and C, with higher relative
abundances on the grains than in the liquors. Lactobacil-
lus nagelii was found in water kefir A, with higher relative
abundances in the liquors than on the grains. Addition-
ally, a low relative abundance of Lactobacillus satsumensis
and Lactobacillus harbinensis was found in the liquors of
water kefirs A and C respectively. Saccharomyces cerevisiae
was the most abundant yeast species in the liquors and
on the grains of water kefirs A, B and C, with a higher
relative abundance on the grains than in the liquors.
Additionally, Zygotorulaspora florentina was found in
water kefir A and Dekkera bruxellensis in water kefirs B
and C, whereby their relative abundances were higher in
the liquors than on the grains.
EPS production was found for 50 and 29% of the L.
hilgardii strains from water kefirs A and C, respectively,
for 48% of the L. nagelii strains from water kefir A, and
for all L. satsumensis strains from water kefir A. None of
the L. paracasei strains from water kefirs A, B and C pro-
duced EPS. The proportions of EPS-producing L. hilgar-
dii and L. nagelii strains were similar in the liquors and
on the grains.
Culture-independent species diversities and community
dynamics
The PCR-DGGE community profiles of the liquors and
grains with the four different primer pairs used were sim-
ilar for the three biological replicates at each sampling
point of water kefirs A, B and C, and were stable as a
function of time during the entire courses of the three
fermentation processes (data not shown). With the V3
and/or LAC primer pairs, gel bands assigned to L. nagelii,
L. hilgardii and L. paracasei/casei were found for water
kefirs A, B and C (Fig. 4). Although it is not straightfor-
ward to correlate band intensities with species abun-
dances, relative comparisons often indicate certain trends.
Indeed, the relative intensities of the bands assigned to L.
nagelii were higher for the liquors than for the grains,
and those of the bands assigned to L. hilgardii were
higher for the grains than for the liquors. The relative
intensities of the bands assigned to L. paracasei/casei were
lower for water kefir A than for water kefirs B and C,
both for the liquors and the grains. Additionally, bands
assigned to Lactobacillus mali/hordei and L. harbinensis
(V3 and LAC primer pairs) and an Oenococcus species
(V3 primer pair) were found in water kefir C. Both
showed higher relative intensities for the liquors than for
the grains. With both the V3 and Bif primer pairs, Bifi-
dobacterium aquikefiri (100% identity; accession no.
LN849254) was found during the entire fermentation
processes of water kefirs A and C, but not in water kefir

(a) (b)
100 100
5 4

80 3 2 80
3

% of total isolates
% of total isolates

60 60

40 40
2

20 1 20
1
0 0
Kefir AL AG BL BG CL CG AL AG BL BG CL CG Kefir
Isolates 35 45 34 51 22 27 35 40 48 64 56 82 Isolates

Figure 3 Culture-dependent microbial species diversities of the liquors (subscript L) and grains (subscript G) of the water kefir fermentation pro-
cesses inoculated with water kefir grain inocula A, B and C, obtained by pooling the samples from different sampling points. The closest known
type strains of the sequenced fragments are given. (a) Bacterial species diversity: 1, Lactobacillus paracasei (100% identity, accession no.
AP012541); 2, L. hilgardii (99% identity, accession no. LC064898); 3, Lactobacillus nagelii (99% identity, accession no. NR112754); 4, Lactobacil-
lus harbinensis (100% identity, accession no. NR028658); and 5, Lactobacillus satsumensis (99% identity; accession no. NR028658). (b) Yeast
species diversity: 1, Saccharomyces cerevisiae (large subunit rRNA gene (LSU) (99% identity, accession no. KC881066)) and internal transcribed
spacer (ITS) region (99% identity, accession no. KC881067)); 2, Dekkera bruxellensis (LSU (100% identity, accession no. AY969049) and ITS
(100% identity, accession no. NR111030)); and 3, Zygotorulaspora florentina (LSU (100% identity, accession no. U72165) and ITS (100% identity,
accession no. AY046168)).

726 Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology
D. Laureys and L. De Vuyst Impact of the water kefir inoculum

V3 LAC diversity data. Indeed, their relative intensities were

AL AG BL BG CL CG AL AG BL BG CL CG higher for the liquors than for the grains, which was in
5 line with the culture-dependent results for Z. florentina.
5
6 Substrate consumption and metabolite production
7
6 profiles
6 1 1
1 1 7 Sucrose was the main carbohydrate at the start of the
4 4
three fermentation processes, and was completely con-
1 1 4 4 sumed (residual concentrations <1 g l 1) after 24, 48 and
1 1 1 1 144 h of fermentation for water kefirs A, B and C respec-
4 1 1 tively (Fig. 2). The glucose concentrations decreased con-
4 tinuously during all three fermentation processes, whereas
3 3 3 3 3 3
1 1 1 the fructose concentrations initially increased to reach a
4 4
2 2 maximum after approximately 24 h of fermentation. The
1 1 TRC concentrations in water kefirs A, B and C were
4 4 4
<1 g l 1 after 48, 72 and 144 h respectively (Table 2).
4 4 4 The time until the TRC concentrations were <1 g l 1 was
higher when the water kefir grain growth was higher and
Figure 4 Culture-independent bacterial community profiles of the was lower when the viable counts of the LAB and yeasts
liquors (subscript L) and the grains (subscript G) of the water kefir fer- on the grains were higher (Tables 2 and 3). Ethanol was
mentation processes inoculated with water kefir grain inocula A, B the most abundant metabolite produced during all three
and C after 48 h of fermentation. The numbers indicate the bands fermentation processes (Table 2). The production of
that were sequenced. The closest known type strains of the ethanol was more or less linear during the first 48 h of
sequenced fragments are given. V3 primer pair: 1, Lactobacillus nage-
fermentation, after which the production slowed down
lii/ghanensis (99% identity; accession no. NR112754/NR043896); 2,
Bifidobacterium aquikefiri (100% identity; accession no. LN849254);
(Fig. 2). This corresponded to the time that glucose was
3, L. hilgardii/diolivorans (100% identity; accession no. LC064898/ depleted in the three fermentation processes. The highest
NR037004); 4, Lactobacillus paracasei/casei/zeae/rhamnosus (100% concentrations of ethanol were found in water kefirs A
identity; accession no. AP012541/AP012544/NR037122/JQ580982); and B, followed by water kefir C. The glycerol production
5, Oenococcus kitaharae (97% identity; accession no. NR041312); 6, paralleled that of ethanol (Fig. 2) and its concentrations
L. mali/hordei (99% identity; accession no. LC064888/NR044394); were similar for the three water kefirs when the TRC
and 7, L. harbinensis (100% identity; accession no. NR113969). LAC
concentrations were <1 g l 1. The ratios of the glycerol
primer pair: 1, L. nagelii (99% identity; accession no. NR119275); 3,
L. hilgardii/diolivorans (99% identity; accession no. LC064898/
to ethanol concentrations were highest in water kefir C
NR037004); 4, L. paracasei/casei/zeae (100% identity; accession no. (Table 2). The highest concentrations of lactic acid were
AP012541/AP012544/NR037122); 6, L. mali/hordei (99% identity; found in water kefir A, followed by water kefirs B and C,
accession no. LC064888/NR044394); and 7, L. harbinensis (100% and the proportion of D-lactic acid was lowest in water
identity; accession no. NR113969). kefir A (Table 2 and Fig. 2). The production of acetic
acid paralleled that of lactic acid, and the highest concen-
B. With the yeast primer pair (data not shown), the trations were found in water kefir C and the lowest in
bands with the highest intensities for the liquors and the water kefir B (Table 2). The concentrations of lactic acid
grains of water kefirs A, B and C were assigned to S. cere- and acetic acid continued to increase after the TRC con-
visiae (100% identity; accession no. KC881066). Their rel- centrations were <1 g l 1 (Fig. 2). The production of
ative intensities were always higher for the grains than for mannitol followed the same pattern as that of the acetic
the liquors. For the liquors and the grains of water kefirs acid production in the three fermentation processes until
B and C, bands with weak relative intensity were assigned the TRC concentrations were <1 g l 1.
to D. bruxellensis (100% identity; accession no. No major aroma compounds were found in the WKSM
AY969049). Their relative intensities were higher for (before inoculation) via SH-GC-MS. The concentrations
water kefir C than for water kefir B, and were higher in of isoamylalcohol, 2-methyl-1-propanol, ethyl hexanoate,
the liquors than on the grains, which confirmed the cul- ethyl octanoate and isoamyl acetate (SH-GC-MS)
ture-dependent results. In water kefir A, bands that could increased fast during the first 48 h of the three fermenta-
not be identified by sequencing were present above the tion processes, whereafter their concentrations remained
band attributed to S. cerevisiae, which might be attributed more or less stable (Fig. 5). In contrast, the concentra-
to Z. florentina, based on the culture-dependent species tions of ethyl acetate and ethyl decanoate increased only

Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology 727
Impact of the water kefir inoculum D. Laureys and L. De Vuyst

80 Isoamyl alcohol (mg l–1) 10 Ethyl decanoate (mg l–1)

60 8
6
40
4
20
2
0 0
0 48 96 144 192 0 48 96 144 192
20 Ethyl octanoate (mg l–1) 60 Ethyl acetate (mg l–1)

15
40
Figure 5 Concentrations of isoamyl alcohol,
10 ethyl octanoate, ethyl decanoate and ethyl
20 acetate (mg l 1) as a function of time (h),
5
measured in the static headspace (SH) of the
0 0 liquors of the water kefir fermentation
0 48 96 144 192 0 48 96 144 192 processes inoculated with water kefir grain
Time (h) Time (h) inocula A (●), B ( ) and C (○).

slowly during the first 24 h of fermentation, whereafter
their concentrations continued to increase until the end
of the fermentation processes. When the TRC concentra-
tions were <1 g l 1, water kefir A contained the highest
and lowest isoamyl acetate and ethyl decanoate concen-
trations, respectively, water kefir B the highest ethyl
octanoate concentrations, and water kefir C the highest
ethyl acetate and lowest ethyl hexanoate concentrations
(Table 4). The concentrations of the esters found in the
liquors of water kefirs A, B and C were always well above
their threshold values when the TRC concentrations were
<1 g l 1. Although some esters, higher alcohols and
short- to long-chain fatty acids were already found in the
WKSM (before inoculation) via SPME-GC-MS, their rela-
tive abundances increased significantly during the three
fermentation processes (Table 5). In contrast, the aldehy-
des hexanal, furfural and benzaldehyde were found in the
WKSM (before inoculation) but not in the liquors after
fermentation. Water kefirs B and C contained higher
relative abundances of 4-ethylphenol and 4-ethylguaiacol
than water kefir A and the WKSM. Water kefir A con-
tained the highest relative abundances of benzyl alcohol
and 2-phenylethyl acetate, and the lowest ones of ethyl 2-
methyl-butanoate, ethyl heptanoate, ethyl nonanoate and
decanoic acid (Table 5). Water kefir B contained the low-
est relative abundances of ethyl benzenepropanoate and
diethyl succinate. Water kefir C contained the highest
relative abundances of ethyl lactate, ethyl 2-methyl-
butanoate, ethyl benzenepropanoate, decanoic acid and
1-octanol, and the lowest ones of ethyl butanoate and
ethyl 9-decenoate. There was no evidence for the presence
of 1,3-propanediol neither in WKSM (before inoculation)
nor in the three liquors after fermentation.
Water kefir grain exopolysaccharide composition
The water kefir grains were completely hydrolysed into
glucose after 6 h of treatment. No other monosaccharides

Table 4 Concentrations of aroma compounds (mg l 1) in the static headspace (SH) of the liquors of the water kefir fermentation processes inoc-
ulated with water kefir grain inocula A, B and C, when the total carbohydrate concentrations were <1 g l 1. Statistically significant differences
(P < 0
05) are indicated with superscript letters

Water kefir fermentation process (time)

Compound KI* Id* Threshold† A (48 h) B (72 h) C (144 h)

2-Methyl-1-propanol 1097 MS/RF 40 16
80  1
28a 10
24  1
06b 14
86  0
80a

Isoamyl alcohol 1222 MS/RF 30 59
50  3
20a 47
61  1
87b 44
74  1
59b
Ethyl acetate 831 MS/RF 7
5 15
03  2
83a 15
00  3
95a 41
29  4
98b
Isoamyl acetate 1141 MS/RF 0
03 0
57  0
08a 0
13  0
03b 0
14  0
02b
Ethyl hexanoate 1250 MS/RF 0
014 0
97  0
12a 0
92  0
11a 0
59  0
06b
Ethyl octanoate 1450 MS/RF 0
005 9
13  0
46a 13
74  2
47b 9
31  2
02a
Ethyl decanoate 1659 MS/RF 0
2 1
37  0
47a 3
14  1
25b 3
48  0
66b

*The Kovats index (KI) and the method of identification (Id) are given for every compound. Identification was via the mass spectrum (MS) and by
comparison with the retention time of the reference compound (RF).
†See references (Corison et al. 1979; Lambrechts and Pretorius 2000; Mamede et al. 2005; Molina et al. 2009).

728 Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology
D. Laureys and L. De Vuyst Impact of the water kefir inoculum

Table 5 Relative abundances of the aroma compounds expressed in arbitrary units (AU) found after solid phase micro-extraction of the head-
space of the water kefir simulation medium (WKSM) and the liquors of the water kefir fermentation processes inoculated with water kefir grain
inocula A, B and C, when the total carbohydrate concentrations were <1 g l 1. Statistically significant differences (P < 0
05) are indicated with
superscript letters

Water kefir fermentation process (time)

Compound KI* Id* WKSM A (48 h) B (72 h) C (144 h)

2-Phenylethyl acetate 1861 MS/RF 0  0a

635  73b
123  14c
89  7c
Ethyl lactate 1386 MS/RF 1  1a 186  24b 118  10b,c 308  89c
Ethyl butanoate 1027 MS 0  0a 64  11b 71  2b 48  2c
Ethyl 2-methyl-butanoate 1045 MS 0  0a 7  2b 19  1c 23  1d
Ethyl heptanoate 1377 MS 0  0a 25  12b 44  4c 47  7c
Ethyl nonanoate 1560 MS 0  0a 41  20b 69  8c 85  18c
Ethyl 9-decenoate 1723 MS 1  1a 171  52b 185  28b 73  15c
Ethyl benzenepropanoate 1928 MS 0  0a 69  8b 49  1c 99  17d
Methyl octanoate 1437 MS 10  1a 21  13a,c 46  4b 31  6c
Isoamyl octanoate 1674 MS 1  0a 59  16b 74  12b,c 88  21c
Diethyl succinate 1702 MS 2  0a 366  2b 145  7c 344  61b
Hexanoate 1870 MS/RF 16  1a 176  27b 241  19c 217  27b,c
Octanoate 2069 MS/RF 18  2a 606  85b 824  112c 710  147b,c
Nonanoate 2170 MS 0  0a 11  8b 12  3b 11  2b
Decanoate 2271 MS 6  1a 90  19b 207  34c 258  67c
Hexanal 1092 MS/RF 32  5a 2  1b 1  0b 1  0b
Furfural 1511 MS/RF 66  4a 10  9b 6  6b 1  1b
Benzaldehyde 1570 MS/RF 41  3a 0  0b 0  0b 0  0b
1-Octanol 1570 MS 0  0a 45  8b 59  2b 91  17c
1,3-Propanediol 1831 MS/RF NF NF NF NF
Benzyl alcohol 1913 MS 1  0a 13
5  2
1b 4
8  0
7a 6
4  8
4a
2-Phenylethanol 1954 MS/RF 3  1a 1058  88b 929  53b 911  125b
4-Ethylphenol 2187 MS/RF 1  1a 5  1b 22  1c 16  2d
4-Ethylguaiacol 2063 MS 6  1a 17  4a 241  22b 180  20c
2,4-di-tert-butylphenol 2102 MS 15  2a 87  9b 162  17c 143  21c
Butylated hydroxytoluene 2310 MS 10  1a 27  5b 36  1c 38  5c
Styrene 1311 MS 1  1a 26  15b 38  1b 39  5b

*The Kovats index (KI) and the method of identification (Id) are given for every compound. Identification was via the mass spectrum (MS) and by
comparison with the retention time of the reference compound (RF).
NF, not found.

or organic acids were found, indicating that they were
composed of glucan-type EPS.
Carbon recovery
The carbon recoveries were approximately 100% during
the entire courses of the three fermentation processes,
indicating that all major substrates and metabolites were
recovered. The values when the TRC concentrations were
<1 g l 1 were representative for the entire courses of the
fermentation processes (Table 2).
Discussion
The integrative multiphasic and comparative approach of
this study allowed to determine the influence of the water
kefir grain inoculum on the characteristics of a water
kefir fermentation process. Additionally, the comparative
nature of this study allowed revealing associations
between the microbial species diversity and fermentation
characteristics such as water kefir grain growth.
The viable counts of the LAB and yeasts, the most
abundant micro-organisms during water kefir fermenta-
tion, remained stable during the entire course of the
three processes studied. Microbial growth resulted from
water kefir grain growth, as the water kefir grains always
harboured the majority of the micro-organisms, confirm-
ing previous results (Laureys and De Vuyst 2014). Fur-
thermore, the time until total carbohydrate exhaustion
was lower when the viable counts of the micro-organisms
on the grains were higher. The stable character of the
viable microbial counts was reflected in a stable microbial
species diversity during the entire courses of the three
processes studied. All three water kefirs harboured
L. paracasei (most prevalent), L. hilgardii, L. nagelii and
S. cerevisiae. These species, regularly reported in the

Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology 729
Impact of the water kefir inoculum
literature on water kefir, may be the key micro-organisms
for water kefir fermentation (Pidoux 1989; Galli et al.
1995; Magalh~aes et al. 2010, 2011; Gulitz et al. 2011,
2013; Miguel et al. 2011; Hsieh et al. 2012; Marsh et al.
2013; Laureys and De Vuyst 2014). Additionally, other
LAB were found in water kefirs A and C, which were
absent in water kefir B. Lactobacillus hilgardii was more
abundant on the grains than in the liquors and when the
water kefir grain growth was high. Indeed, isolated strains
produced EPS from sucrose, as shown before (Pidoux
et al. 1988, 1990; Waldherr et al. 2010). However, the
mere presence of EPS-producing strains of L. hilgardii
was not sufficient for good water kefir grain growth, as
shown by water kefir A. Lactobacillus paracasei was more
abundant when the water kefir grain growth was high,
although none of the strains isolated from the water
kefirs of the present study produced EPS from sucrose, in
contrast with strains from a previous study (Gulitz et al.
2011). Hence, L. paracasei was probably not responsible
for water kefir grain growth. The relative abundances of
L. nagelii were inversely related with water kefir grain
growth, even though some strains isolated from water
kefir A produced EPS from sucrose. Furthermore, this
micro-organism was more abundant in the liquors than
on the grains and thus probably not responsible for water
kefir grain growth. The novel species B. aquikefiri (Lau-
reys et al. 2016) was found in water kefirs A and C but
not in water kefir B, indicating that it was not necessary
for water kefir grain growth neither for the course of the
fermentation process.
Low water kefir grain growth was associated with small
grains with high viable counts. The brittle water kefir
grains break easily during sieving or handling, and insuf-
ficient water kefir grain growth may cause them to
become small gradually. Small grains have a large specific
surface and can harbour high viable counts, as the micro-
organisms are mainly attached onto the surface of the
grains (Moinas et al. 1980; Laureys and De Vuyst 2014).
This further explained why low water kefir grain growth
was associated with fast fermentation. Water kefir grain
growth resulted from the partial conversion of sucrose
into glucan EPS by extracellular glucansucrases (Monsan
et al. 2001). The activity of glucansucrases decreases at
low pH values (Waldherr et al. 2010), so the low pH val-
ues during fermentation process A may have caused its
low water kefir grain growth. However, the pH did not
drop so fast to exclude any glucansucrase activity, making
it more likely that the production of glucansucrases by
L. hilgardii was suppressed by the low pH values. Acidic
stress may thus cause low water kefir grain growth, which
should be investigated in more detail. When the water
kefir grain growth decreased, less glucose was incorpo-
rated into grain EPS and hence more glucose remained
730 Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology
D. Laureys and L. De Vuyst
available for acid production, further increasing the acidic
stress. Over multiple backsloppings, a continuous increase
in the acidic stress may result in a continuous decrease of
the water kefir grain growth, as was seen during the
prefermentations of the present study. This illustrated
that it will be necessary to adjust the process parameters
based on the fermentation characteristics to maintain a
stable water kefir production process.
Glucose was the preferred substrate during the water
kefir fermentation processes studied, as it was always con-
sumed faster than fructose. Ethanol, lactic acid, glycerol,
acetic acid and mannitol were the main end-metabolites
produced. Despite a stable ratio of LAB to yeast cells of
2–10, the majority of the metabolites was always pro-
duced by the yeasts. The production of mannitol indi-
cated the use of fructose as alternative external electron
acceptor by heterofermentative LAB (Zaunm€ uller et al.
2006), but the mannitol concentrations remained low,
confirming previous results (Laureys and De Vuyst 2014).
Part of the acetate production may be attributed to bifi-
dobacteria, as higher acetate concentrations in water
kefirs A and C coincided with their presence (Laureys
et al. 2016). Nevertheless, given the low acetate concen-
trations in these water kefirs, bifidobacterial metabolism
was of minor impact during fermentation. Continued
bacterial metabolism in all the water kefir fermentation
processes studied after carbohydrate depletion may be
ascribed to the fermentation of other (not measured) car-
bohydrates, such as starch derived from the figs or glu-
cans composing the grains. Bifidobacteria were probably
not the main cause of this extended metabolism, as they
were absent in water kefir B. Although there was no evi-
dence for the degradation of the grain EPS, dextranase
activity has already been shown in certain bifidobacterial
(Kaster and Brown 1983) and LAB species (Picozzi et al.
2015).
Isoamyl acetate, ethyl hexanoate, ethyl octanoate and
ethyl decanoate possess fruity and floral aromas and may
exert an important influence on the aroma of the water
kefir liquors, as their concentrations were much higher
than their threshold values (Lambrechts and Pretorius
2000). In contrast, the concentrations of ethyl acetate,
isoamyl alcohol and 2-methyl-1-propanol were around
their threshold values. These may contribute a harsh and
unpleasant solvent-like aroma at high concentrations, but
may add desirable complexity to fermented beverages in
lower concentrations (Lambrechts and Pretorius 2000).
Free fatty acids (sour, cheesy, sweaty, rancid, soapy and/
or goaty aroma), short- to medium-chain esters (fruity
and floral), long-chain esters (soapy) and 2-phenylethanol
(rosy) were produced, whereas hexanal, furfural and ben-
zaldehyde disappeared during all three fermentation
processes studied (Vandermerwe and Vanwyk 1981;
D. Laureys and L. De Vuyst
Lambrechts and Pretorius 2000). The compounds 4-ethyl-
phenol (wet horse) and 4-ethylguaiacol (smoky, vanilla
and clove-like) are associated with the metabolism of
D. bruxellensis, and their abundances were indeed higher
when this yeast species was present during fermentation
(Lambrechts and Pretorius 2000). The absence of 1,3-pro-
panediol indicated that glycerol was not further converted
by LAB species such as L. hilgardii (Pasteris and de Saad
2009; Bauer et al. 2010).
In conclusion, this comparative study allowed to deter-
mine the key micro-organisms from the wide range of
microbial species found in water kefir, namely L. paraca-
sei, L. hilgardii, L. nagelii and S. cerevisiae. The presence
of EPS-producing L. hilgardii strains was not sufficient
for good water kefir grain growth. Low water kefir grain
growth seemed to be caused by low pH values. Further-
more, the water kefir grain growth seemed to impact the
size of the grains, which may in turn impact the fermen-
tation rate. This study will be of value for the selection of
an appropriate water kefir grain inoculum and for devel-
oping and maintaining a stable water kefir production
process.
Acknowledgements
The authors acknowledge their financial support of the
Research Council of the Vrije Universiteit Brussel (SRP7,
IRP2, and IOF342 projects) and the Hercules Foundation
(grant UABR09004). D.L. was the recipient of a PhD fel-
lowship of the Vrije Universiteit Brussel.
Conflict of Interest
No conflict of interest declared.
References
Altschul, S.F., Gish, W., Miller, W., Myers, E.W. and Lipman,
D.J. (1990) Basic local alignment search tool. J Mol Biol
215, 403–410.
Bauer, R., du Toit, M. and Kossmann, J. (2010) Influence of
environmental parameters on production of the acrolein
precursor 3-hydroxypropionaldehyde by Lactobacillus
reuteri DSMZ 20016 and its accumulation by wine
lactobacilli. Int J Food Microbiol 137, 28–31.
Corison, C.A., Ough, C.S., Berg, H.W. and Nelson, K.E.
(1979) Must acetic acid and ethyl acetate as mold and rot
indicators in grapes. Am J Enol Vitic 30, 130–134.
Edwards, U., Rogall, T., Blocker, H., Emde, M. and Bottger,
E.C. (1989) Isolation and direct complete nucleotide
determination of entire genes – characterization of a gene
coding for 16S-ribosomal RNA. Nucleic Acids Res 17,
7843–7853.
Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology
Impact of the water kefir inoculum
Emaga, T.H., Rabetafika, N., Blecker, C.S. and Paquot, M.
(2012) Kinetics of the hydrolysis of polysaccharide
galacturonic acid and neutral sugars chains from flaxseed
mucilage. Biotechnol Agron Soc Environ 16, 139–147.
Galli, A., Fiori, E., Franzetti, L., Pagani, M.A. and Ottogalli, G.
(1995) Microbiological and chemical composition of sugar
kefir grains. Ann Microbiol Enzimol 45, 85–95.
Gulitz, A., Stadie, J., Wenning, M., Ehrmann, M.A. and Vogel,
R.F. (2011) The microbial diversity of water kefir. Int J
Food Microbiol 151, 284–288.
Gulitz, A., Stadie, J., Ehrmann, M.A., Ludwig, W. and Vogel,
R.F. (2013) Comparative phylobiomic analysis of the
bacterial community of water kefir by 16S rRNA gene
amplicon sequencing and ARDRA analysis. J Appl
Microbiol 114, 1082–1091.
Hsieh, H.H., Wang, S.Y., Chen, T.L., Huang, Y.L. and Chen,
M.J. (2012) Effects of cow’s and goat’s milk as
fermentation media on the microbial ecology of sugary
kefir grains. Int J Food Microbiol 157, 73–81.
Janssens, M., Myter, N., De Vuyst, L. and Leroy, F. (2012)
Species diversity and metabolic impact of the microbiota
are low in spontaneously acidified Belgian sausages with
an added starter culture of Staphylococcus carnosus. Food
Microbiol 29, 167–177.
Kaster, A.G. and Brown, L.R. (1983) Extracellular dextranase
activity produced by human oral strains of the genus
Bifidobacterium. Infect Immun 42, 716–720.
Lambrechts, M.G. and Pretorius, I.S. (2000) Yeast and its
importance to wine aroma: a review. S Afr J Enol Vitic 21,
97–129.
Laureys, D. and De Vuyst, L. (2014) Microbial species
diversity, community dynamics, and metabolite kinetics of
water kefir fermentation. Appl Environ Microbiol 80, 2564–
2572.
Laureys, D., Cnockaert, M., De Vuyst, L. and Vandamme, P.
(2016) Bifidobacterium aquikefiri sp. nov. isolated from
water kefir. Int J Syst Evol Microbiol 66, 1281–1286.
Leroy, F., Vasilopoulos, C., Van Hemelryck, S., Falony, G. and
De Vuyst, L. (2009) Volatile analysis of spoiled, artisan-
type, modified-atmosphere-packaged cooked ham stored
under different temperatures. Food Microbiol 26, 94–102.
Magalh~aes, K.T., Pereira, G.V.D., Dias, D.R. and Schwan, R.F.
(2010) Microbial communities and chemical changes
during fermentation of sugary Brazilian kefir. World J
Microbiol Biotechnol 26, 1241–1250.
Magalh~aes, K.T., Pereira, G.V.D., Campos, C.R., Dragone, G.
and Schwan, R.F. (2011) Brazilian kefir: structure,
microbial communities and chemical composition. Braz J
Microbiol 42, 693–702.
Mamede, M.E.O., Cardello, H.M.A.B. and Pastore, G.M.
(2005) Evaluation of an aroma similar to that of sparkling
wine: sensory and gas chromatography analyses of
fermented grape musts. Food Chem 89, 63–68.
Marsh, A.J., O’Sullivan, O., Hill, C., Ross, R.P. and Cotter,
P.D. (2013) Sequence-based analysis of the microbial
731
Impact of the water kefir inoculum
composition of water kefir from multiple sources. FEMS
Microbiol Lett 348, 79–85.
Miguel, M.G.D.P., Cardoso, P.G., Magalhaes, K.T. and Schwan,
R.F. (2011) Profile of microbial communities present in
tibico (sugary kefir) grains from different Brazilian states.
World J Microbiol Biotechnol 27, 1875–1884.
Moens, F., Lefeber, T. and De Vuyst, L. (2014) Oxidation of
metabolites highlights the microbial interactions and role
of Acetobacter pasteurianus during cocoa bean
fermentation. Appl Environ Microbiol 80, 1848–1857.
Moinas, M., Horisberger, M. and Bauer, H. (1980) The
structural organization of the tibi grain as revealed by
light, scanning and transmission microscopy. Arch
Microbiol 128, 157–161.
Molina, A.M., Guadalupe, V., Varela, C., Swiegers, J.H.,
Pretorius, I.S. and Agosin, E. (2009) Differential synthesis
of fermentative aroma compounds of two related
commercial wine yeast strains. Food Chem 117, 189–195.
Monsan, P., Bozonnet, S., Albenne, C., Joucla, G., Willemot,
R.M. and Remaud-Simeon, M. (2001)
Homopolysaccharides from lactic acid bacteria. Int Dairy J
11, 675–685.
Papalexandratou, Z., Falony, G., Romanens, E., Jimenez, J.C.,
Amores, F., Daniel, H.M. and De Vuyst, L. (2011) Species
diversity, community dynamics, and metabolite kinetics of
the microbiota associated with traditional Ecuadorian
spontaneous cocoa bean fermentations. Appl Environ
Microbiol 77, 7698–7714.
Pasteris, S.E. and de Saad, A.M.S. (2009) Sugar-glycerol
cofermentations by Lactobacillus hilgardii isolated from
wine. J Agric Food Chem 57, 3853–3858.
Picozzi, C., Meissner, D., Chierici, M., Ehrmann, M.A.,
Vigentini, I., Foschino, R. and Vogel, R.F. (2015) Phage-
mediated transfer of a dextranase gene in Lactobacillus
sanfranciscensis and characterization of the enzyme. Int J
Food Microbiol 202, 48–53.
Pidoux, M. (1989) The microbial flora of sugary kefir grain
(the gingerbeer plant): biosynthesis of the grain from
Lactobacillus hilgardii producing a polysaccharide gel. J
Appl Microbiol 5, 223–238.
Pidoux, M., Brillouet, J.M. and Quemener, B. (1988)
Characterization of the polysaccharides from a
Lactobacillus brevis and from sugary kefir grains. Biotechnol
Lett 10, 415–420.
732 Journal of Applied Microbiology 122, 719--732 © 2016 The Society for Applied Microbiology
D. Laureys and L. De Vuyst
Pidoux, M., Deruiter, G.A., Brooker, B.E., Colquhoun, I.J. and
Morris, V.J. (1990) Microscopic and chemical studies of a
gelling polysaccharide from Lactobacillus hilgardii.
Carbohyd Polym 13, 351–362.
Rimaux, T., Vrancken, G., Vuylsteke, B., De Vuyst, L. and
Leroy, F. (2011) The pentose moiety of adenosine and
inosine is an important energy source for the fermented-
meat starter culture Lactobacillus sakei CTC 494. Appl
Environ Microbiol 77, 6539–6550.
Vandermerwe, C.A. and Vanwyk, C.J. (1981) The contribution
of some fermentation products to the odor of dry white
wines. Am J Enol Vitic 32, 41–46.
Vilgalys, R. and Hester, M. (1990) Rapid genetic identification
and mapping of enzymatically amplified ribosomal DNA
from several Cryptococcus species. J Bacteriol 172, 4238–
4246.
Waldherr, F.W., Doll, V.M., Meissner, D. and Vogel, R.F.
(2010) Identification and characterization of a glucan-
producing enzyme from Lactobacillus hilgardii TMW 1.828
involved in granule formation of water kefir. Food
Microbiol 27, 672–678.
White, T., Buns, T., Lee, S. and Taylor, J. (1990) Amplification
and direct sequencing of fungal ribosomal RNA genes for
phylogenetics. In PCR Protocols: A Guide to Methods and
Applications ed. Innis, N., Gelfand, D., Snisky, J. and
White, T. pp. 315–322. NY, USA: Academic Press.
de Winter, J.C.F. (2013) Using the Student’s t-test with
extremely small sample sizes. Pract Ass Res Eval 18, 1–12.
Wouters, D., Bernaert, N., Anno, N., Van Droogenbroeck, B.,
De Loose, M., Van Bockstaele, E. and De Vuyst, L.
(2013a) Application and validation of autochthonous
lactic acid bacteria starter cultures for controlled leek
fermentations and their influence on the antioxidant
properties of leek. Int J Food Microbiol 165, 121–133.
Wouters, D., Bernaert, N., Conjaerts, W., Van Droogenbroeck,
B., De Loose, M. and De Vuyst, L. (2013b) Species
diversity, community dynamics, and metabolite kinetics of
spontaneous leek fermentations. Food Microbiol 33, 185–
196.
Zaunm€ uller, T., Eichert, M., Richter, H. and Unden, G. (2006)
Variations in the energy metabolism of biotechnologically
relevant heterofermentative lactic acid bacteria during
growth on sugars and organic acids. Appl Microbiol
Biotechnol 72, 421–429.