To assess the structure and stability of a dominant lactic acid bacteria (LAB) population during the propagation of rye sourdough in an industrial semi-fluid production over a period of 7 months.
To assess the structure and stability of a dominant lactic acid bacteria (LAB) population during the propagation of rye sourdough in an industrial semi-fluid production over a period of 7 months.
The sourdough was started from a 6-year-old freeze-dried sourdough originating from the same bakery. A unique microbial consortium consisting mainly of bacteria belonging to species Lactobacillus helveticus, Lactobacillus panis and Lactobacillus pontis was identified based on culture-dependent (Rep-PCR) and culture-independent [denaturing gradient gel electrophoresis (DGGE)] methods. Three of the isolated Lact. helveticus strains showed remarkable adaptation to the sourdough conditions. They differed from the type strain by the ability to ferment compounds specific to plant material, like salicin, cellobiose and sucrose, but did not ferment lactose.
We showed remarkable stability of a LAB consortium in rye sourdough started from lyophilized sourdough and propagated in a large bakery for 7 months. Lactobacillus helveticus was detected as the dominant species in the consortium and was shown to be metabolically adapted to the sourdough environment.
The use of an established and adapted microbial consortium as a starter is a good alternative to commercial starter strains.
The taste, aroma and texture of sourdough breads greatly depend on starter cultures used for flour fermentation (Salim-ur-Rehman et al. 2006; Corsetti and Settanni 2007). Rye sourdough cultures are mainly lactic acid bacteria (LAB) belonging to both homo- and heterofermentative species. So far, the composition of microbial communities of rye sourdoughs used in different countries has been characterized mainly based on phenotypical traits (reviewed Hansen 2004; De Vuyst and Neysens 2005). Lactobacillus brevis, Lactobacillus plantarum and Lactobacillus sanfranciscensis are the species that have most frequently been isolated from rye sourdoughs. In addition, Lactobacillus amylovorus, Lactobacillus panis, Lactobacillus reuteri (Rosenquist and Hansen 2000) and Lactobacillus fermentum (Weckx et al. 2010) have been reported as dominating species in some rye sourdough consortia.
Several factors, including process parameters, production environment and type of flour, can affect the microbial composition of sourdoughs (De Vuyst and Neysens 2005; Scheirlinck et al. 2007). Spontaneous fermentation of flour is commonly used to make sourdough in traditional bread making. Microbial consortia with remarkably high stability develop in such sourdoughs during continuous propagation. This kind of associations may endure propagation in the form of back-slopping for years, in spite of nonaseptic conditions (De Vuyst and Neysens 2005). According to the study by Van der Meulen et al. (2007), it may take as short as 10 days to establish a stable LAB consortium, but commonly it takes much longer.
The tendency in the manufacture of sourdough breads is to initiate fermentation by adding defined commercial starter cultures with specific properties (De Vuyst and Neysens 2005). Unfortunately, commercial strains are often not competitive with endogenous microflora entering the sourdough back-slopping process from flour and the environment of the bakery (Meroth et al. 2003; Siragusa et al. 2009). Consequently, frequent renewing of the sourdough cycling process is required. In our work, we determined the composition of a lyophilized rye sourdough and monitored the dynamics of the microbial population in an industrial sourdough made from the lyophilized starter. The lyophilized sourdough aliquots had been prepared from sourdough used for bread making in the same bakery 6 years ago. The microbial community was monitored during 7 months of daily propagation using both culture-dependent and culture-independent approaches.
The sourdough was propagated in a large bakery (800 kg of sourdough) in Estonia. The lyophilized starter used for the inoculation of the new sourdough was prepared from the sourdough of the same bakery 6 years before. It was stored at −20°C in hermetically closed packages. The renewal of the industrial sourdough cycle was started by mixing lyophilized starter with rye flour and tap water in the ratio of 10 : 36 : 54 and incubating for 24 h at 32°C. The following four back-slopping renewals were performed in the same ratio after 19, 12, 12 and 10 h of fermentation. Routine industrial back-slopping was carried out with the same mother sponge, rye flour and water ratio of 10 : 36 : 54 (sourdough yield 250 kg dough per 100 kg flour) and fermentation at 32°C for 10 h. Samples for chemical and microbiological analysis were collected at the end of fermentations.
To measure pH and total titratable acidity (TTA) of fermented sourdough, five grams of sample was suspended in 50 ml of distilled water. pH and TTA were measured with Food and Beverage Analyzer D22 (Mettler-Toledo International Inc., Columbus, OH, USA). The pH value was recorded, and the acidity was determined by titration using 0·1 N NaOH to final pH 8·5. TTA was expressed as ml 0·1 N NaOH per 10 g of sourdough. All measurements were performed in triplicate.
The number of colony forming LAB in sourdough samples was determined by plate counting. Five grams of sourdough and 45 ml of 0·85% NaCl solution were added to a sterile 50-ml centrifuge tube and mixed on a vortex for 5 min. Series of decimal dilutions were plated on MRS agar (LabM, Lancashire, UK) in triplicate. Plates were incubated at 30°C in both aerobic and anaerobic conditions for 48 h. The BD GasPak EZ System (Becton Dickinson Microbiology Systems, Franklin Lakes, NJ, USA) was used to create anaerobic environment.
The Whatman indicating FTA MiniCards (GE Healthcare Ltd., UK) were used for DNA extraction from isolated colonies as described by manufacturer. DNA extracted with this method was used for Rep-PCR analysis.
A modified phenol–chloroform method described by Van der Meulen et al. (2007) and Camu et al. (2007) was used to extract DNA directly from sourdough samples and pure LAB cultures for denaturing gradient gel electrophoresis (DGGE) analysis and 16S rRNA gene sequencing. The bacterial pellet obtained from 5 g of sourdough or 1·5 ml of overnight culture was washed in 1 ml of TES buffer [6·7% sucrose, 50 mmol l−1 Tris–HCl (pH 8·0), 50 mmol l−1 EDTA] and resuspended in 300 μl STET buffer [8% sucrose, 5% Triton X–100, 50 mmol l−1 Tris–HCl (pH 8·0) and 50 mmol l−1 EDTA]. A volume of 75 μl of TES lysis buffer containing 1170 U ml−1 mutanolysin and 100 mg ml−1 lysozyme and 100 μl of proteinase K (2·5 mg ml−1) (all enzymes from Sigma-Aldrich Co. LLC., St Louis, MO,USA) was added, and samples were incubated at 37°C for 1 h. After incubation, 40 μl of preheated (37°C) 20% SDS in TE buffer and a pinch of glass beads (diameter 150–212 μm) were added. Samples were mixed on vortex for 1 min and incubated at 37°C for 10 min, followed by incubation at 65°C for 10 min. Two phenol–chloroform–isoamylalcohol (50 : 49 : 1) extractions and one chloroform–isoamylalcohol (49 : 1) extraction were carried out. DNA was precipitated from the aqueous phase by adding 0·1 volumes of 3 mol l−1 sodium acetate and two volumes of cold 96% ethanol and washed with 70% ethanol. The pellet was dried at room temperature and resuspended in 50 μl of TE buffer (10 mmol l−1 Tris–Cl, 1 mmol l−1 EDTA, pH 8·0; AppliChem GmbH, Darmstadt, Germany).
Rep-PCR (repetitive element palindromic PCR) with primer (GTG)5 was performed essentially as described by De Vuyst et al. 2002. For each sourdough sample, 40 colonies were analysed by Rep-PCR. Twenty colonies were picked from one aerobically incubated plate and another 20 colonies from one anaerobically incubated plate. Colony picking was performed successively from one sector of the plate. PCR was carried out in 25-μl volume using the following cycle: preliminary denaturation 6 min at 95°C; amplification in 30 cycles: denaturation 94°C for 1 min, annealing 1 min at 40°C, extension 8 min at 65°C and final extension at 65°C for 16 min. Share of each LAB fingerprint type among aerobically and anaerobically selected isolates was calculated as the ratio of similar fingerprints to the number of analysed colonies.
One to two representatives of each Rep-PCR fingerprint group were subjected to 16S rRNA gene analysis. 16S rRNA gene fragments were amplified using universal primers 27f-YM (Frank et al. 2008) and 16R1522 (Weisburg et al. 1991) followed by column purification of the amplified fragment with GeneJET PCR Purification kit (Fermentas, Vilnius, Lithuania) and set up for sequencing PCR using BigDye Terminator v3.1 Cycle Sequencing kit as described by manufacturer (Applied Biosystems, Foster City, CA, USA). The partial 16S rRNA gene sequences obtained (approx. 700 bp) were compared with GenBank database using the blast algorithm (National Center for Biotechnology Information, USA).
For DGGE analysis, DNA was amplified using primers F357GC and 518R as described by Gafan and Spratt (2005). DNA was extracted directly from the sourdough samples, pure cultures of identified LAB isolates and 10 kGy irradiated rye flour. Sterile dark rye flour type R1370 (Tartu Grain Mill Ltd., Tartu, Estonia) was obtained by γ-irradiation at 10 kGy using a dosimetric system GEX WinDose (Centennial, CO, USA). Polyacrylamide gel (8% acrylamide-N,N′-methylenebisacrylamide; 37·5 : 1) with a gradient from 35 to 70% of urea and formamide (100% corresponding to 40% v/v formamide and 7 mol l−1 urea) was used. Electrophoresis was performed with the INGENYphorU system (Ingeny BV International, goes the Netherlands) at constant voltage 75 V and temperature 60°C for 18 h. The gels were stained with ethidium bromide and photographed with the ImageQuant 400 system (GE Healthcare, Little Chalfont, UK). Bands of interest were excised and DNA eluted by incubating in TE buffer overnight at 4°C. Eluted DNA was reamplified using same primers without the GC-clamp. DNA fragments obtained were cloned using InsTAclone PCR Cloning kit (Fermentas) combined with the TransformAid Bacterial Transformation kit (Fermentas). Cloned DNA fragments were sequenced using the M13 primer.
Metabolic profiles of Lact. helveticus strains were determined with API 50 CH testing, and proteolytic activity was studied with the API ZYM test (Humble et al. 1977) as described by the manufacturer (bioMérieux, Marcy l'Etoile, France). Skim milk agar was used to assess the ability of the bacteria to hydrolyse casein (Marcy and Pruett 2001). The test was performed at both 30 and 37°C in aerobic and anaerobic conditions. Lactobacillus helveticus CH-1 (Chr. Hansen, Hørsholm, Denmark) was used as a reference strain.
The number of colony forming LAB in lyophilized starter was 2 × 106 and 107 CFU g−1 in aerobic and anaerobic conditions, respectively. The counts were approximately two magnitudes lower compared with the freshly fermented sourdough (Table 1). This may indicate that freeze-drying and storage during 6 years had a negative effect on the cultivability of LAB.
|Number of back-slopping propagations||Time from the start of renewal||pH||TTA ml 0·1 N NaOH per10 g sourdough||Aerobic LAB counts CFU g−1||Anaerobic LAB counts CFU g−1|
|0||0 h||4·79 ± 0·11||4·98 ± 0·43||8·91 ± 0·23 × 104||7·08 ± 0·07 × 105|
|1||24 h||3·63 ± 0·06||20·74 ± 0·48||1·95 ± 0·02 × 108||7·59 ± 0·04 × 108|
|2||43 h||3·53 ± 0·14||21·27 ± 0·15||4·27 ± 0·00 × 109||1·10 ± 0·01 × 109|
|3||55 h||3·54 ± 0·14||17·66 ± 0·09||2·40 ± 0·03 × 108||6·46 ± 0·08 × 108|
|4||67 h||3·54 ± 0·12||17·34 ± 0·38||2·82 ± 0·03 × 108||1·07 ± 0·01 × 109|
|5||77 h||3·66 ± 0·13||15·71 ± 0·27||1·29 ± 0·00 × 108||5·62 ± 0·07 × 108|
|40a||2 weeks||ND||ND||1·17 ± 0·05 × 108||ND|
|100a||5 weeks||3·67 ± 0·06||22·30 ± 0·56||4·57 ± 0·02 × 107||6·61 ± 0·05 × 108|
|250a||14 weeks||3·71 ± 0·01||18·89 ± 0·05||4·68 ± 0·06 × 107||1·10 ± 0·01 × 109|
|300a||16 weeks||3·63 ± 0·02||19·78 ± 0·07||1·29 ± 0·00 × 108||1·41 ± 0·01 × 109|
|350a||18 weeks||3·63 ± 0·04||21·30 ± 0·41||1·29 ± 0·00 × 108||6·92 ± 0·05 × 108|
|550a||34 weeks||3·53 ± 0·02||21·84 ± 0·30||3·31 ± 0·03 × 108||1·20 ± 0·00 × 109|
|600a||35 weeks||3·49 ± 0·02||21·28 ± 0·08||2·19 ± 0·02 × 108||1·41 ± 0·02 × 109|
The industrial sourdough cycle was started by inoculation with the freeze-dried sourdough at final LAB concentration of 106 CFU g−1 dough (Table 1). The number of LAB in the sourdough increased to 109 CFU g−1 already 24 h after inoculation. During the following 7 months, the number of LAB stayed around 109 CFU g−1 (Table 1). Colony forming numbers were generally 10 times lower on plates incubated in aerobic conditions (Table 1).
Acidity of the renewed sourdough quickly increased to 20 ml 0·1 N NaOH per 10 g sourdough during the first 24 h of incubation (Table 1). The acidity slightly decreased after third- to fifth-back-slopping step, presumably due to decreased fermentation time. After 5 weeks of back-slopping, the pH and TTA values of the industrial sourdough were around 3·5–3·7 and 19–22, respectively (Table 1).
The microbial composition of sourdough samples was evaluated using Rep-PCR fingerprinting with (GTG)5 primer. Altogether 40 LAB isolates were analysed from each sample, 20 of which were picked from aerobically and 20 from anaerobically incubated plates. Three to six fingerprint types per sample were detected among LAB isolated in anaerobic conditions. LAB isolated in aerobic conditions were represented by one to three fingerprint types per sample. Fingerprint types of the aerobic isolates were not unique and coincided with those of anaerobic isolates (data not shown).
One or more representatives from each fingerprint group (Fig. 1) were subjected to partial sequencing of the 16S rRNA gene (Table S1). Bacteria belonging to Lact. helveticus, Lact. panis, Lactobacillus pontis and Lactobacillus vaginalis species were identified among isolates obtained from the lyophilized starter. Most of the randomly picked isolates belonged to species Lact. helveticus irrespective of incubation conditions (Table 2). Species identified from the lyophilized starter were represented by one (Lact. helveticus N92 and Lact. vaginalis N1113) or two (Lact. panis N915, N1311 and Lact. pontis N139, N131) fingerprint types (Fig. 1, Table S2).
|LAB species||Lyophilized||77 h||5 weeks||14 weeks||18 weeks||34 weeks||35 weeks|
Representatives of Lact. helveticus species were the dominating aerobic isolates obtained from industrial rye sourdough samples during the 7-month-long observation period. They formed 17/20 to 20/20 of the isolates in aerobic conditions (Table 2). The subdominant population varied among the samples and consisted of LAB belonging to species Lact. vaginalis, Lact. fermentum or Lactobacillus paralimentarius.
Lactobacillus helveticus was also the dominating colony forming species among LAB isolated in anaerobic conditions. However, its proportion among colony forming bacteria gradually decreased as it was replaced by bacteria belonging to species Lact. pontis and Lact. panis (Table 2). Representatives of Lact. vaginalis and Lactobacillus casei were sporadically detected among the subdominant population.
Three fingerprint patterns were detected for Lact. helveticus species (N720, N92 and E96), two fingerprint patterns for Lact. panis (N915 and N1311), Lact. pontis (N139 and N131) and Lact. vaginalis (N1113 and N116) species, and one fingerprint pattern for Lact. casei (N726), Lact. fermentum (E112) and Lact. paralimentarius (E712) species (Fig. 1). Lactobacillus helveticus bacteria with N92 and E96 Rep-PCR profiles were isolated in both aerobic and anaerobic conditions, and Lact. helveticus N720 fingerprint type was from a single isolate from sourdough sample obtained 5 weeks after the start of the new sourdough cycle. Only Lact. helveticus N92 bacteria were isolated from the lyophilized starter, and they were also the dominating type among all Lact. helveticus isolates in the industrial sourdough during 7 months of propagation (Table S2).
The culture-independent PCR-DGGE analysis was used in parallel to plating and Rep-PCR fingerprinting to characterize changes in the microbial consortium of sourdough. Preliminary PCR-DGGE analysis of the isolated LAB revealed that bacteria identified as Lact. helveticus, Lact. vaginalis and Lact. fermentum species produced a single separately migrating band on the gel. Bacteria belonging to Lact. panis, Lact. pontis and Lact. casei species, however, showed multiple bands, possibly due to the heterogeneity of 16S rRNA gene operons (Fig. 2). The presence of one band in the control sample (10 kGy irradiated flour) indicated that plant genome can be amplified with the universal primers F357-GC and 518R. Sequencing analysis of this band showed similarity to mitochondrial cereal DNA.
PCR-DGGE fingerprinting of the freeze-dried sourdough starter and industrial sourdough samples revealed a remarkable similarity in band patterns during the whole 7-month period. Bands with positions corresponding to the dominating LAB species Lact. helveticus, Lact. pontis and Lact. panis were observed in all fingerprinting patterns. The PCR-DGGE band patterns also showed no major changes in the microbial population of the industrial sourdough during the 7 months of propagation. Only the detectable appearance of Lact. vaginalis and decrease in Lact. panis as a result of daily propagation could be suggested based on the DGGE analysis of the industrial sourdough (Fig. 2). Identity of all bands was additionally confirmed by sequencing analysis.
16S rRNA gene sequencing and PCR-DGGE analysis determined Lact. helveticus as dominating species in the lyophilized starter and industrial sourdough. However, Lact. helveticus is used as a starter in the dairy industry and is not generally known as a sourdough LAB. API 50 CH test was used to evaluate metabolic potential of the Lact. helveticus isolates in parallel to Lact. helveticus CH-1 strain used in the dairy industry.
Three Lact. helveticus strains (Lact. helveticus N720, Lact. helveticus N92 and Lact. helveticus E96) representing different fingerprint groups were studied. All three strains isolated in this study could ferment glucose, esculin ferric citrate, saccharose, salicin and cellobiose (Table 3). Maltose and mannose were fermented by Lact. helveticus E96 and N720, but not by the N92 strain representing dominating fingerprinting group of the sourdough. The Lact. helveticus N720 also fermented galactose and N-acetyl glucosamine. None of the three Lact. helveticus isolated from the rye sourdough could ferment lactose, contrary to Lact. helveticus CH-1 strain, which fermented lactose, mannose, fructose, glucose and galactose.
|Carbohydrate in API 50CH||Lact. helveticus N720||Lact. helveticus N92||Lact. helveticus E96||Lact. helveticus CH-1||Lact. helveticus profile APIa|
|Esculin ferric citrate||+||+||+||−||−|
We determined the proteolytic activity of Lact. helveticus N92, E96 and N720 strains and the reference dairy strain Lact. helveticus CH-1 using skim milk agar test and API ZYM test panel. Skim milk agar test showed that only the reference strain CH-1 was able to hydrolyse casein (Table 4). None of the strains had trypsin or α-chymotrypsin activity. The activity of specific arylamidases that catalyse the hydrolysis of N-terminal amino acid from peptides, amides or arylamides differed among the tested strains (Table 4).
|Enzyme tested||Lact. helveticus N720||Lact. helveticus N92||Lact. helveticus E96||Lact. helveticus CH-1|
Lactobacillus helveticus was the dominant LAB species in the lyophilized starter and industrial sourdough samples. Unlike Lact. panis (Wiese et al. 1996) and Lact. pontis (Vogel et al. 1994) also detected among dominant LAB population of the studied sourdough, Lact. helveticus is not a common sourdough LAB. It is often found in fermented milk products and is used as a cheese starter culture (Slattery et al. 2010; Broadbent et al. 2011). So far, Lact. helveticus has been isolated from Sudanese sorghum sourdough (Hamad et al. 1997), traditional wheat or wheat–rye sourdoughs of East-Flanders region in Belgium (Scheirlinck et al. 2008) and some commercial starters (Moroni et al. 2010). The metabolic profiles of Lact. helveticus isolated from sourdough in our work significantly differed from those determined for type strain and the dairy starter.
Blast analysis of partial 16S rRNA gene sequences of isolates N92, E96 and N720 (GenBank ID: HM641233, HM623785 and HM641232, respectively) first determined Lactobacillus suntoryeus as the closest match (data not shown). Lactobacillus suntoryeus is a recently identified species (Cachat and Priest 2005), which afterwards was claimed to be a later synonym of Lact. helveticus (Naser et al. 2006). Representatives of Lact. suntoryeus species were first isolated from late stages of barley fermentation in Japanese and Scottish malt whisky distilleries (Cachat and Priest 2005) and similarly to Lact. helveticus belong to thermophilic homofermentative lactobacilli (Cachat and Priest 2005).
In spite of the high similarity of Lact. helveticus and Lact. suntoryeus at sequence level, which determined them as one species, they formed distinguishable groups based on SDS-PAGE protein profiles and metabolic profiling (Naser et al. 2006). Lactobacillus helveticus type strains are able to ferment lactose and/or galactose (Sharpe 1981; Naser et al. 2006; API test identification table; this work). In contrast, none of the strains isolated from whisky distilleries (and primarily identified as Lact. suntoryeus) could ferment lactose (Naser et al. 2006). Instead they were able to ferment disaccharides cellobiose and sucrose, compounds found in plant material, and salicin, which is a β-glucoside produced from bark. Similarly, Lact. helveticus strains isolated in our work were able to utilize cellobiose, sucrose and salicin, but not lactose or galactose. The only exception was Lact. helveticus N720, which additionally utilized galactose, thereby resembling another metabolically versatile Lact. helveticus strain R0052 (Naser et al. 2006).
Lactobacillus helveticus are known to have a vast proteolytic system containing several proteinases that are important for the technological potential of these bacteria in the dairy industry (Christensen et al. 1999). Rapid growth in milk relies on complex proteolytic system, whose collective function involves the release of essential amino acids from large proteins such as casein (Christensen et al. 1999). However, none of the Lact. helveticus strains we isolated from sourdough were able to hydrolyse casein.
Assumingly genomic rearrangements have occurred in the adaptation process of Lact. helveticus to specific environments. Lactobacillus helveticus genome encompasses more than 200 insertion sequence (IS) elements (Callanan et al. 2008) that are responsible for the heterogeneity among Lact. helveticus strains (Kaleta et al. 2010) and have been associated with truncations in genes associated with cellobiose transport (Callanan et al. 2007) and the lac gene cluster (Callanan et al. 2005). Sequencing of Lact. helveticus N92, N720 and E96 genomes is in progress to resolve the adaptation mechanisms to the sourdough environment.
Lactobacillus helveticus remained dominating in the microbial population of industrial sourdough during 7 months of daily back-slopping. However, its proportion among cultivable population decreased over time. Unlike the culture-dependent method, DGGE analysis did not reveal significant changes in the microbial composition of industrial sourdough or in the species ratio. All three dominant species (Lact. helveticus, Lact. panis and Lact. pontis) and also Lact. vaginalis were detected with DGGE analysis. Representatives of Lact. vaginalis were also detected in the lyophilized starter used for renewing the sourdough cycle.
Culture-independent methods, like DGGE, have several advantages compared with culture-dependent methods in the study of sourdough, because LAB from continuously cycled sourdough can have low colony forming ability (De Vuyst et al. 2002; Iacumin et al. 2009). Also, media used for the isolation of sourdough lactobacilli quantitatively and qualitatively influence the microbial population that is detected (Vera et al. 2009).
Although same LAB fingerprint types were isolated from aerobically and anaerobically incubated agar plates, the colony numbers were generally 10 times higher on anaerobically grown plates. Additional experiments revealed that most of the isolated LAB were sensitive to oxygen (Mihhalevski et al. 2011). The industrial sourdough studied is characterized by high water content and large volumes (up to 800 kg), which limit the specific transport rate of oxygen from dough surface. Reversible sorption of oxygen by dough starch and gluten leads to low levels of dissolved oxygen in flour water slurry (Xu 2001). Thus, the microbial consortium has adapted to semi-anaerobic conditions.
Development of the microbial community in sourdough greatly depends on cycling conditions: fermentation temperature, type of flour, back-slopping rate and frequency (Hammes et al. 1996; Vogel et al. 1996; Hammes and Gänzle 1998). Commercial starters may not be sufficiently adapted to cycling conditions and are consequently outcompeted by indigenous bacteria entering the sourdough cycle from the environment. However, if the strains are obtained from a cycle with similar parameters, like in this study, the consortium can be stable even after several months of back-slopping. Thus, using an established and adapted microbial consortium as a starter could be a good option for bakeries.
The financial support for this research was provided by the European Regional Development Fund project EU29994, Estonian Ministry of Education and Research grant SF140090s08, Estonian Science Foundation grant ETF9417 and by European Social Fund's Doctoral Studies and Internationalization programme DoRa.