Impact of process conditions on the microbial community dynamics and metabolite production kinetics of teff sourdough fermentations under bakery and laboratory conditions

Abstract Teff and teff sourdoughs are promising ingredients for bread production. Therefore, this study aimed at the characterization of spontaneous and flour‐native starter culture‐initiated teff sourdough productions under bakery and laboratory conditions. Backslopped laboratory and bakery teff sourdough productions were characterized by different lactic acid bacteria (LAB) and yeast species, but were both characterized by a pH below 4.0 after five backslopping steps. The sourdough‐associated Lactobacillus sanfranciscensis was isolated for the first time from backslopped spontaneous teff sourdoughs. The autochthonous strain L. sanfranciscensis IMDO 150101 was tested as starter culture during laboratory teff sourdough fermentations. Its prevalence could be related to the process conditions applied, in particular the ambient temperature (below 30°C). Breads made with 20% teff sourdough (on flour basis) displayed interesting features compared with all‐wheat‐based reference breads. Teff sourdoughs were characterized as to their pH evolution, microbial community dynamics, and microbial species composition. Representative strains of the LAB species isolated from these sourdoughs, in particular L. sanfranciscensis, may be selected as starter cultures for the production of stable teff sourdoughs and flavorful breads, provided they are adapted to the environmental conditions applied.

These LAB species mainly produce lactic acid (homo-and heterofermenters) and acetic acid (heterofermenters). The yeasts Candida glabrata and Wickerhamomyces anomalus have commonly been found in laboratory wheat sourdoughs, whereas Saccharomyces cerevisiae has commonly been found in laboratory barley sourdoughs Vrancken et al., 2010).
Whereas different parts of the kernel can be fractionated in the case of wheat, rye, and barley, the whole kernel is used in the case of teff because of its small size. This may influence the fermentation process because of the presence of bran Katina, Laitila et al., 2007;Prückler et al., 2015). Teff contains many proteins (providing all essential amino acids, including lysine), slowly digestible complex carbohydrates (causing satiety), many fibers (improving gut health), and more bioavailable minerals (among which calcium and iron) (Ashenafi, 2006;Gebremariam et al., 2014). These properties make teff an interesting product for human consumption in general and for consumption by elderly, veganists, and sport men in particular.
The aim of this study was to determine the LAB and yeast species diversity, microbial community dynamics, and metabolite production kinetics of spontaneous teff sourdough fermentations performed through backslopping under bakery and laboratory conditions and to assess the competitiveness of L. sanfranciscensis IMDO 150101 as starter culture strain for teff sourdoughs, to be able to develop stable teff sourdoughs for bread production.

| Spontaneous backslopped sourdough productions
The bakery sourdough productions are referred to as TF1′ (flour A), TF2′ (flour A), and TF3′ (flour B), and the laboratory sourdough productions are referred to as TF1 (flour A) and TF2 (flour B).

| Starter culture-initiated sourdough productions
Starter culture-initiated laboratory sourdough fermentations were carried out with the L. sanfranciscensis IMDO 150101 strain, which was isolated from backslopped teff sourdough production TF1′ during this study. Both small-scale fermentations in glass bottles (350 mL, dough yield of 400) and fermentor-scale fermentations (8 L, dough yield of 400) were carried out. The former were performed to assess the impact of the temperature on the survival and prevalence of the L. sanfranciscensis strain used, whereas the latter allowed a comparison of teff fermentations with laboratory nonteff sourdough fermentations carried out before Van Kerrebroeck, Bastos, Harth, & De Vuyst, 2016;Vrancken et al., 2011;Weckx, Van der Meulen, Allemeersch et al., 2010;Weckx, Van der Meulen, Maes et al., 2010). The flour-water mixtures were inoculated with a cell suspension of the starter culture strain at a final concentration of 10 6 -10 7 colony forming units (CFUs)/mL. The glass bottles were shaken at 160 revolutions per minute (rpm) to prevent sedimentation of flour particles and were incubated at 23°C, 30°C, or 37°C for 120 hr. The temperature of the fermentors was kept constant at 30°C for 72 hr; the mixture was kept homogeneous through stirring at 300 rpm. All starter culture-initiated fermentations were performed in triplicate. They are further referred to as TFSC23, TFSC30, and TFSC37 (flour B, small-scale fermentations) and as TFFS1, TFFS2, and TFFS3 (flour C, fermentor-scale fermentations).

| Sampling and analyses
Sampling procedure, determination of pH and total titratable acidity (TTA), culture-dependent (plating on different agar media, from which colonies were picked up) and culture-independent microbial community dynamics [denaturing gradient gel electrophoresis of targeted PCR amplicons from sample DNA, PCR-DGGE; LAB, all sourdoughs (and flours); yeasts; all sourdoughs (and flours), except for the small-scale fermentations), classification and identification of LAB [all sourdoughs (and flours); (GTG) 5 -PCR fingerprinting] and yeast isolates [backslopped sourdoughs (and flours); M13-PCR fingerprinting], and metabolite target analyses for all sourdough productions were carried out as described previously ).

| Culture-dependent analysis
The agar media used were de Man-Rogosa-Sharpe-5 (mMRS-5) agar medium (Meroth, Walter, Hertel, Brandt, & Hammes, 2003), supplemented with cycloheximide (final concentration of 0.1 g/L; Sigma-Aldrich, Saint Louis, MO, USA) for LAB isolation (including the L. sanfranciscensis IMDO 150101 strain) and yeast extractpeptone-glucose (YPG) agar medium , chloramphenicol being present in a final concentration of 0.1 g/L, for yeast isolation. Incubations were performed at 30°C for 48 to 72 hr. Also, samples from the flour batches were analyzed microbiologically. Therefore, 10 g of flour was mixed with 10 mL of saline (0.85% NaCl, m/v), a tenfold dilution series of these suspensions was made, and 100 μL of each dilution was plated on mMRS-5 and YPG agar media supplemented with cycloheximide or chloramphenicol in a final concentration of 0.1 g/L, respectively.

| Metabolite target analysis
The concentrations of glucose, fructose, sucrose, maltose, and mannitol were determined by high-performance anion exchange chromatography with pulsed amperometric detection, those of ethanol and acetic acid by gas chromatography with flame ionization detection, and those of lactic acid by high-performance liquid chromatography with refractive index detection, as described previously . Volatile compounds were determined qualitatively through gas chromatography coupled to mass spectrometry in conjunction with solid-phase microextraction of the sourdough headspace (HS/SPME-GC-MS), as described previously .

| Statistical analysis of volatile compound data
A principle component analysis (PCA) was performed on the peak area data of the HS/SPME-GC-MS volatile analysis. This was followed by a cluster analysis, using the software package SPSS 20.0 (SPSS, Chicago, IL, USA). To determine the number of principal components (PCs), a scree plot was constructed. To maximize the sum of the squares of the correlations between the original variables and the rotated PCs (factor loadings), a Varimax with Kaiser normalization rotation was applied. A three-dimensional score plot was constructed.

| Bread production and evaluation
Teff sourdough-based breads were produced in the pilot plants of four industrial bakeries according to their respective recipes and breadmaking conditions, with addition of 20% (m/m, on flour basis) teff sourdough from the backslopped (after backslopping step 10) sourdough productions (bakery sourdough productions TF1′, TF2′, and TF3′, and laboratory sourdough productions TF1 and TF2), and from the starter culture-initiated sourdough production TFFS1. In addition to, all-wheat-based reference breads (without the addition of teff sourdough) were produced. Parbaked breads were produced at the last fermentation day and baked off prior to evaluation. The breads were assessed by 21 consumers on the basis of descriptive data.

| Backslopped bakery sourdough productions
During the first two backslopping steps of the spontaneous backslopped bakery teff sourdough productions TF1′and TF2′, no significant change in pH and TTA occurred, whereas the pH dropped from 5.9 to 4.6 and the TTA increased from 3.5 to 11.4 mL for backslopped bakery teff sourdough production TF3′ (Figure 1a). After the third backslopping step, the pH of TF1′ and TF2′ dropped from 5.9 and 5.6 to 4.4 and 4.8, respectively, and further down to 3.9 and 4.1 at the end of both backslopping processes. In the case of TF3′, the pH continuously dropped to reach a final value of 3.7.
Both bakery teff sourdough productions TF1′ and TF2′ showed a continuous increase of the TTA from the third backslopping step onwards. Bakery sourdough production TF3′ showed an increase of the TTA value to 28.0 mL after five backsloppings. At the end of the backslopping processes for TF1′ and TF2′, TTA values of 21.3 and 14.9 mL, respectively, were reached. The TTA of the teff sourdough production TF3′ continuously dropped from the fifth backslopping step onwards to a final value of 23.1 mL at the end of the backslopping process.

| Backslopped laboratory sourdough productions
The pH of the spontaneous backslopped laboratory teff sourdough productions TF1 and TF2 decreased from 6.1 to 3.9 and from 6.0 to 4.6 after the first 24 hr of fermentation, respectively, while the TTA values increased from 2.4 to 14.3 mL and from 3.2 to 13.8 mL, respectively. During these laboratory sourdough productions, both the pH and TTA evolved faster compared with the bakery ones ( Figure 1b). After seven backsloppings, the pH reached an average value of 3.7 and remained more or less stable upon further backslopping for both processes, while the TTA values increased toward an average value of 18.1 and 13.1 mL for the respective backslopping processes. A pH of 3.5 and 3.7 was reached at the end of the respective backslopping processes, which corresponded with TTA values of 17.4 and 12.3 mL.
The initial LAB counts (10 4 CFU/g) of backslopped bakery teff sourdough production TF3′ were higher and increased faster, whereas the yeast counts were the same but increased faster, compared with TF1′ and TF2′. The LAB counts of TF1′ and TF2′ F I G U R E 1 Evolution of pH (lines) and total titratable acidity (TTA; bars) during teff sourdough productions carried out under bakery conditions (a; TF1′, light gray; TF2′, dark gray; TF3′, black) and laboratory conditions (b; TF1, light gray; TF2, dark gray) reached values of approximately 10 9 CFU/g after the fourth backslopping step, whereas the yeast counts reached values of approximately 10 7 CFU/g after the seventh backslopping step and remained stable upon further backslopping. In the case of TF3′, high LAB counts of 10 9 CFU/g were reached after the first backslopping step, which remained more or less stable until the end of the backslopping process. The yeast counts of TF3′ reached values of >10 7 CFU/g after the second backslopping step and remained constant until the end of the backslopping process.
During each of the backslopped bakery teff sourdough productions, more than 100 LAB isolates were obtained from mMRS-5 agar media, the identities of which are represented in Table 1.
During TF1′, L. sanfranciscensis, W. cibaria/confusa, and L. sakei were present from the beginning of the backslopping process, with an increasing relative abundance of L. sanfranciscensis and a decreasing one of L. sakei upon backslopping ( Figure 3a). Weissella cibaria/confusa was outcompeted after the fourth backslopping step, whereas L. coryniformis grew out from then on. During TF2′, P. acidilactici and W. confusa were outcompeted by L. helveticus, which was dominant from the second backslopping step (Figure 3b). Lactobacillus helveticus represented 77% of all isolates, albeit that it was not isolated from the other backslopped bakery teff sourdough productions. In the case of TF3′, L. brevis was isolated throughout the whole backslopping process and became the dominant LAB species at the end of the backslopping process, although accompanied with P. pentosaceus that was present from backslopping step 7 ( Figure 3c). Lactococcus lactis decreased in relative abundance from the start till backslopping step 2, whereas L. coryniformis was present till the fourth backslopping step.

| Backslopped laboratory sourdough productions
Low LAB counts of 10 3 CFU/mL and no yeast counts were found at the start of both backslopped laboratory teff sourdough productions TF1 and TF2 ( Figure 2c). These LAB counts reached values of >10 9 CFU/mL after the first backslopping step, whereas the yeast counts reached values of ≥10 7 CFU/mL after the second backslopping step (Figure 2c,d). Afterward, the LAB counts remained stable.
However, the yeast counts were lower during the fifth, sixth, and seventh backslopping steps.
More than 200 colonies were picked up from mMRS-5 agar media from plated samples of the backslopped laboratory teff sourdough productions TF1 and TF2, the identities of which are represented in Table 1

| Starter culture-initiated laboratory sourdough fermentations
Concerning the impact of the fermentation temperature on the prevalence of L. sanfranciscensis IMDO 150101 as added starter culture strain for teff sourdoughs, the small-scale fermentations showed that this strain was able to prevail in the fermentations performed at 23°C and 30°C (Table 1). At 30°C, L. sanfranciscensis IMDO 150101 prevailed until 48 hr of fermentation. However, it was outcompeted by L. fermentum and P. acidilactici after 72 hr of fermentation. At 37°C, L. fermentum, P. acidilactici, and P. pentosaceus were the prevailing LAB species.
Prior to inoculation of the starter culture-initiated fermentorscale teff sourdough fermentations, low initial LAB counts (3.0 log CFU/mL) and even lower initial yeast counts (<3.0 log CFU/mL) were present. After 24 hr of fermentation, stable LAB counts between 8.0 and 9.7 log CFU/mL were found, except for TFFS3, in which they declined. The added starter culture strain did not prevail during TFFS1 and TFFS3, but it did during TFFS2. Instead, the former fermentations were characterized by the prevalence of L. fermentum and W. cibaria/confusa, respectively. The latter species was retrieved either as the sole LAB species (TFFS3) or in combination with other LAB species (TFFS1 and TFFS2). After 72 hr of fermentation, yeast counts between 5.5 and 7.0 log CFU/mL were found.

| Backslopped laboratory sourdough productions
Culture-independent analysis based on 16S rRNA-PCR-DGGE bacterial community profiling revealed three different phases during the backslopped laboratory teff sourdough productions TF1 and TF2 ( Figure 4d,e). The first phase (first two backslopping steps) of TF1 was  Culture-independent analysis based on 26S rRNA-PCR-DGGE fungal community profiling revealed the occurrence of S. cerevisiae throughout the whole backslopped laboratory teff sourdough production TF2 (Figure 4g). No PCR amplicons could be generated for samples of TF1.

| Starter culture-initiated laboratory sourdough productions
16S rRNA-PCR-DGGE bacterial community profiling revealed that L. sanfranciscensis IMDO 150101 as added starter culture strain did not prevail during two of the three teff fermentor-scale sourdough fermentations carried out (Figure 5a-c). It did prevail during TFFS2, in combination with W. cibaria/confusa. The latter species was found in most fermentations, including TFSC30 and TFSC37.
Initially, other fungal species were retrieved as well, in particular species of plant pathogens. In the case of TFFS1 and TFFS2, one band of PCR amplicons could not be identified (Figure 5d,e).

No other yeast species could be identified during TFFS1, whereas
Cyberlindnera fabianii was detected during TFFS2. In all cases, plant DNA was found too.
F I G U R E 5 Culture-independent bacterial (a-c) and fungal (d-f) community dynamics through 16S rRNA-PCR-and 26S rRNA-PCR-DGGE (universal primers) analysis during Lactobacillus sanfranciscensis IMDO 150101-initiated fermentor-scale laboratory teff sourdough productions (a,d, TFFS1; b,e, TFFS2; c,f, TFFS3); I represents the inoculum, 0 represents the initial flour-water mixture, 0′ represents the sourdough sample after addition of the starter culture strain; 8, 12, 24, 48, and 72 represent the sourdough samples taken after 8, 12, 24, 48, and 72 hr of fermentation, respectively. Plate washes are represented by an asterisk at the right of the respective time points.

| Backslopped bakery sourdough productions
The unfermented dough mixtures of the three backslopped bak-
Mannitol was produced from the beginning of both backslopping processes and reached maximal concentrations of 53.9 ± 4.1 mM and 23.5 ± 5.3 mM after the eight and fourth backslopping steps, respectively, followed by a decrease toward 37.9 and 0 mM at the end of both sourdough productions. Lactic acid was the main metabolite produced during both TF1 and TF2; its concentration increased steadily to reach maximal concentrations of 177.8 ± 4.3 mM and 142.5 ± 4.4 mM at the end of the respective backslopping processes (Figure 7d,e). The second main metabolite produced was ethanol, which increased until the second and fourth backslopping steps during TF1 (170.2 ± 3.1 mM) and TF2 (99.6 ± 1.6 mM), respectively, followed by a decrease upon further backslopping, to reach concentrations of 111.6 ± 2.1 mM and 63.9 ± 1.1 mM, respectively, at the end of the backslopping processes. Acetic acid was produced in low and moderate concentrations during TF1 and TF2, respectively, with maxima of on average 18.5 mM and 33.8 mM during the whole backslopping processes.

| Starter culture-initiated laboratory sourdough productions
The

| Bread production and evaluation
Based on the descriptive results from the 21 consumers involved, the organoleptic properties of the wheat-based breads made with teff sourdoughs showed differences compared with the all-wheatbased reference breads. The breads made with mature laboratory and bakery teff sourdoughs showed a decrease in volume compared to the reference breads. Independently of the sourdoughs used, the teff sourdough-based breads showed a darker crust and crumb and possessed an acid, grain-like, and malty crumb taste. The sourdough breads made with the bakery teff sourdoughs further possessed distinct flavor compounds.

| D ISCUSS I ON
The use of teff flour and teff sourdough is very limited; its most known application is to prepare injera in Ethiopia and Eritrea (Ashenafi, 2006;Gebremariam et al., 2014). However, given its highquality protein, slowly digestible complex carbohydrates, many fibers, and bioavailable minerals, it is very promising for use in human food production, in particular for bread production (Campo et al., 2016).
During these fermentations, it showed a high adaption to the specific flour used and fermentation conditions applied (Viiard et al., 2013). TF1′: 6, unfermented; 7, after five backslopping steps; 8, after ten backslopping steps; TF2′: 9, unfermented; 10, after five backslopping steps; 11, after ten backslopping steps; and TF3′: 12, unfermented; 13, after five backslopping steps; 14, after ten backslopping steps] and laboratory conditions [green; TF1: 1, unfermented; 2, after five backslopping steps; 3, after ten backslopping steps; and TF2: 4, after five backslopping steps; 5, after ten backslopping steps] capacity of teff flour due to its high ash content, and hence the dominance of mainly heterofermentative LAB species; in turn, they imply high acid tolerance of the dominating LAB and yeast species . Furthermore, batch variations in dominating LAB and yeast species are common during sourdough productions and reflect not only the impact of isolation and identification procedures but also the quality of the flour in terms of age, microbial load, stability, etc., as well as environmental contamination (De Vuyst et al., 2014Huys, Daniel, & De Vuyst, 2013;Minervini et al., 2015). For instance, under the laboratory conditions applied in the study of Moroni et al. (2011), no common LAB species were found in two types of sourdough produced from the same flour either. One type of teff sourdoughs was dominated by two LAB species only, that is, P. pentosaceus and Le. holzapfelii, and the yeast species K. barnettii. A second type of teff sourdoughs harbored several lactobacilli and the yeast species C. glabrata. However, unique isolations such as Le. holzapfelii may be accidental occurrences (De Vuyst et al., 2014. Furthermore, given the production conditions of teff, batchto-batch variations are possibly larger than with commercial wheat flour (Gebremariam et al., 2014). Teff kernels are milled as a whole and, hence, a larger and possibly more variable microbial contamination can be expected (Berghofer, Hocking, Miskelly, & Jansson, 2003;Gebremariam et al., 2014). This was reflected in the presence of plant pathogens at the start of certain of the teff sourdough fermentations performed.
Most of the LAB species found in the mature teff sourdoughs of the present study have been isolated from sourdoughs before (De Vuyst et al., 2014, including teff sourdoughs (Desiye & Abegaz, 2013;Moroni et al., 2010Moroni et al., , 2011. pontis was not retrieved from the teff sourdoughs produced during the present study, although this LAB species was considered to be able to dominate teff sourdough fermentations (Moroni et al., 2010(Moroni et al., , 2011. It is however not commonly present in spontaneous sourdough fermentations, probably because of its limited carbohydrate utilization pattern (Vogel et al., 1994). Lactobacillus brevis, considered to be able to dominate teff sourdough fermentations as well (Moroni et al., 2010), did occur in some backslopped bakery teff sourdoughs of the present study. Its presence may reflect the lower ambient temperature and higher carbohydrate content, in particular fructose, of the backslopped bakery sourdoughs (De Vuyst et al., 2002). Furthermore, the high glucose concentrations, owing to the lack of β-amylase activity in teff flour, could have contributed to the occurrence of Weissella species, as these species preferably grow on glucose (Galle, Schwab, Arendt, & Gänzle, 2010;Gänzle, 2014;Gebremariam, Zarnkow, & Becker, 2013;Gebremariam et al., 2014;Moroni et al., 2011). Additionally, degradation of phenolic compounds has been demonstrated for L. brevis, pediococci, and weissellas (Filannino, Gobbetti, De Angelis, & Di Cagno, 2014). The yeast species K. exigua and S. cerevisiae are commonly associated with sourdough fermentations (Lhomme et al., 2016;Vrancken et al., 2010). Whereas K. exigua (maltose-negative) is part of a mutualistic interaction with L. sanfranciscensis and/or L. brevis (both maltose-positive) in stable sourdoughs, this is not considered to be the case for S. cerevisiae (Gobbetti, 1998;Kline & Sugihara, 1971;Lhomme et al., 2016;Ottogalli, Galli, & Foschino, 1996). Other yeast species usually dominate in the absence of S. cerevisiae or Kazachstania spp. (Moroni et al., 2011;Vrancken et al., 2010). Based on all these and former data, it can be con-  Liu et al., 2016;Minervini et al., 2012;Zhang et al., 2015). Yet, it has been isolated from rye sourdoughs (Kitahara et al., 2005;Spicher, 1984;Spicher & Lönner, 1985;Spicher & Schröder, 1978) and spelt sourdoughs (Scheirlinck et al., , 2008 too. However, L. sanfranciscensis could not be retrieved from diverse wheat, rye, spelt, and barley sourdoughs backslopped in the laboratory Van der Meulen et al., 2007;Vrancken et al., 2011;Weckx, Van der Meulen, Allemeersch et al., 2010;Weckx, Van der Meulen, Maes et al., 2010;Weckx et al., 2011) neither was it competitive in backslopped sorghum sourdoughs (Sekwati-Monang, Valcheva, & Gänzle, 2012). The fermentation temperature and dough yields applied were probably responsible for this. Further, it has been speculated that the presence of phenolic compounds in the flour from grains of C4-plants, such as teff and sorghum, inhibits the growth of L. sanfranciscensis (Gänzle, 2014;Sekwati-Monang et al., 2012). Yet, the retrieval of L. sanfranciscensis from spontaneous teff sourdoughs and its prevalence in certain laboratory teff sourdoughs during the present study underlines that its selection is rather steered by technological process parameters, such as temperature and pH, as well as its preference for maltose as energy source and its association with maltose-negative yeasts, such as Candida humilis (now reclassified as Kazachstania humilis) and K. exigua (De Vuyst et al., 2014. Its presence in the flour and the occurrence of the right environmental and processing conditions for its growth will hence determine its prevalence in sourdoughs spontaneously developed. However, the microbial load of the flour will depend on agricultural crop practices (use of manure), weather conditions (sunshine and rainfall influencing mold growth and hence causing microbial competition), and its contact with rodents and insects throughout the chain from cereal to bread (Boiocchi et al., 2017;Groenewald, Van Reenen, & Dicks, 2006;Minervini et al., 2015;Thaochan, Drew, Chinajariyawong, Sunpapao, & Pornsuriya, 2015). Given the milling of the small teff seeds as a whole and their high surface/volume ratio (Gebremariam et al., 2014;Tefera & Belay, 2006), a high microbial load together with a variable microbial species diversity can be expected in the case of teff flour. The recovery of L. sanfranciscensis from insect frass indicates a probable contamination route of teff flour with L. sanfranciscensis, a species that is typical for sourdoughs but rarely recovered elsewhere (Boiocchi et al., 2017;De Vuyst et al., 2017). Alternatively, berries and flowers might be the carrier for L. sanfranciscensis (Ripari, Gänzle, & Berardi, 2016).
The teff sourdoughs of the present study were rich in volatile compounds and had an impact on the flavor of the breads produced thereof. In general, sourdoughs harbor a wide spectrum of volatile compounds, which originate from enzymatic and chemical reactions involving flour substrates (e.g., aldehydes and alcohols from lipid oxidation) and both bacterial and yeast metabolism (e.g., alcohols, aldehydes, ketones, carboxylic acids, and esters from enzymatic conversions) (Gänzle, 2014). The differences found between sourdoughs produced under laboratory and bakery fermentation conditions only partially aligned with the differences found between liquid and firm sourdoughs , possibly due to the different process conditions applied during the sourdough productions presented here and the differences in microbial species prevailing.

| CON CLUS IONS
Backslopped teff sourdoughs are characterized by different microbial species, based on the process conditions applied during their production. This was illustrated by the prevalence of different LAB species in bakery and laboratory teff sourdoughs produced under different environmental conditions, and the prevalence of a L. sanfranciscensis strain in bakery and laboratory teff sourdoughs fermented at low temperature (<30°C). In general, the use of representative strains of L. fermentum, L. brevis, L. sanfranciscensis, pediococci, and weissellas, adapted to the environmental conditions they will be confronted with, will contribute to the production of stable teff sourdoughs and flavorful baked goods produced thereof.

ACK N OWLED G M ENTS
The authors acknowledge their financial support of the Research Council of the Vrije Universiteit Brussel (SRP7 and IOF342 projects), the Hercules Foundation (project UABR09004), and Flanders' FOOD (projects INNOCEREAL and INNOCEREAL II), including the mills and bakeries involved.

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

E TH I C A L S TATEM ENT
This study did not involve human or animal testing. Consumers (21) were involved to assess the breads produced.