Introduction

Tchoukoutou is an African opaque sorghum beer obtained by mixed fermentation including different yeasts and lactic acid bacteria. This traditional beer is present in all African countries and is also called pito or burukutu in Ghana and Nigeria, kefir beer in South Africa, pombe in Tanzania and “Seven days” beer in Zambia [1, 2]. African sorghum beer is very rich in calories, B-group vitamins (thiamine, folic acid, riboflavin, nicotinic acid) and amino acids such as lysine [3]. Present in popular bars and a dedicated local market, sorghum beer is a refreshing drink much appreciated by a large number of consumers. In different local Tchoukoutou markets and Cabaret (beer gardens), consumers discriminate producers by the organoleptic quality (flavor and taste) and alcohol content of their beer. As observed by Kayode et al. [4] in Benin and Schoustra et al. [5] in Zambia, producers and consumers claim that the traditional sorghum beer offers health benefits such as preventing and curing diarrhea. The health benefits as well as good aroma and flavor of Tchoukoutou are due to the combination of yeast and the LAB content in this traditional starter. The LAB content may also cooperate with yeasts for the bio-protection of the beer by producing organic acid and bacteriocins [6]. During the mashing processes of producing Tchoukoutou and other African beers, the sorghum mash pH is decreased by spontaneous LAB acidification. This bio-acidification, if efficient, should improve amylase, protease, endo-β-glucanase and phosphatase activities [7, 8], lautering performance [9], organoleptic qualities, mineral bioavailability [10] and vitamin production [11].

African sorghum beers are not standardized, and in addition to the variation of organoleptic qualities, several safety problems limit the marketability of these drinks. Indeed, Tchoukoutou, because of its instability as well as short shelf life resulting from uncontrolled and unlimited fermentation, is only consumed in the areas where it is produced. Furthermore, African sorghum beer is consumed while it is still fermenting, and the drink contains many fragments of insoluble materials, like undigested starch and protein, which still remain in the beer. This is likely due to the inappropriate gelatinization time, temperature and the rudimentary materials (basket and cloth) used for filtration. The growth of mold during germination is also a permanent exogenous risk that compromises the safety of African sorghum beer. Indeed, aflatoxin produced by Aspergillus flavus is the most frequently occurring mycotoxin identified in South African traditional beer [12], in Nigeria’s opaque beer [13] and in Malawi’s traditional maize beer [14]. In addition to this exogenous contaminant, the raw material used during beer processing, i.e., sprouted grains of sorghum bicolor, contains a cyanogenic glycoside [15–18]. If this secondary metabolite of sorghum grain remains in the beer, it generates poisonous hydrogen cyanide by the action of enzymes of the intestinal microbiota after consumption by humans [19]. Because of these safety problems, which hinder the marketability of African sorghum beer, a controlled mixed fermentation may be a potential key to upgrading this traditional beer. This challenge may help producers to move from an ancestral, neolithic household process to industrial production using the endogenous Tchoukoutou microorganisms to guarantee its particular organoleptic quality. This will solve the safety problem, increase the shelf life and make this beer as competitive as Berliner Weißbier, Belgian lambic beer and American coolship ales, which are also obtained by mixed fermentation including yeast and LAB. For this purpose, selecting performant yeast and LAB that produce aryl-ß-D-glucosidase is a good approach. Through this approach, dhurrin (sorghum cyanogenic glucoside) and other sorghum glucose conjugate compounds are hydrolyzed thereby contributing to sorghum beer detoxification and flavor enhancement, respectively. To achieve this goal, the first challenge is the identification of different yeasts and LAB which occur during this spontaneous mixed fermentation. In this study, following a polyphasic approach, the different yeasts in the traditional starter used for Tchoukoutou production are isolated, discriminated and classified based on their phenotype, proteins and DNA sequence analysis. Furthermore, based on their metabolic profile, yeasts with interesting biotechnological properties are selected.

Materials and methods

Sample collection, yeast isolation and purification

Thirty-six samples of Tchoukoutou starter were collected in 12 localities of two agroecological zones with high production of African sorghum beer in Benin. Yeast strains were isolated from the starter using the standard serial method. A hundred microliters of the dilution (10−4–10−8) was plated using the spread plate method on oxytetracycline glucose yeast agar (yeast extract 5 g/l, glucose 20 g/l, agar 15 g/l). The selectiveness of the medium was improved by adding 1 % of oxytetracycline per liter of OGYA medium for LAB inhibition. The inoculated plates were incubated at 27 °C for 72 h. The pure yeast colonies were obtained by picking single colonies and streaking them on the same medium following the quadrant streaking technique. The plates were incubated at 27 °C for 48 h.

Phenotypical characterization

Morphological and physiological characterization

Two hundred and forty pure yeast isolates were characterized on the basis of their morphological and physiological characteristics following the methods described by Yarrow [20]. The morphological and physiological data of the online CBS database (www.cbs.knaw.nl) and the key procedure described by Back [21] were used for yeast identification. The results of this phenotypical characterization were labeled as negative (−), positive (+) or weak (w), and R software was used for heat map construction.

MALDI-TOF analysis based on yeast proteome mass spectrum

A hundred and forty-five yeasts, selected from 240 yeasts which were isolated and phenotypically characterized, were submitted to MALDI-TOF test using the ethanol extraction method. After the growth of the different yeast strains on YPG agar for 48 h at 27 °C, a single yeast colony was suspended in 300 µL of ultrapure water, and then, 900 µL of absolute ethanol was added. The proteome of the precipitated cell material was extracted with formic acid and acetonitrile. One microliter of this yeast extract supernatant was dropped on the target, allowed to dry, overlaid with 1 µL of a matrix solution (α-cyano-4-hydroxy-cinnamic acid). The measurement was performed on a Microflex LT spectrometer (Bruker Daltonik). The Bruker database (Biotyper 3.1) and an in-house database (Technische Mikrobiologie, Weihenstephan) were used for the automatic online yeast identification. According to the similarity between the applied yeast proteome mass spectrum and the mass spectrum of the target yeast strain in the database, the score value is generated. The identification is highly probable at species level when the score value is between 2.300 and 3.000. No reliable identification is observed when the score value is below 1.699. The applied yeast strain being scored from 1.700 to 2.000 limits the identification to genus level. On the basis of the yeast mass spectra, groupings were visualized by similarity calculations and multidimensional scaling (MDS) proposed by Usbeck et al. [22].

Molecular characterization

The overnight culture was obtained by growing the yeast strain in YPG medium at 27 °C. The DNA of this overnight culture was isolated according to the method described by Hanna and Xiao [23].

Amplification of ITS region and RFLP analysis

The PCR was performed with ITS1 (5′-TCCGTAGGTGAACCTGCGG-3′) and ITS4 (5′-TCCTCCGCTTATTGATATGC-3′) primers to amplify the ITS region [24]. This amplified DNA of 50 yeasts which were characterized on the basis of MALDI-TOF analysis was randomly selected from different sub-clusters obtained after multidimensional scaling (MDS). It was digested with three endonucleases HindIII, HinfI and HaeIII, as per the instruction of the manufacturer (New England Biolabs, Frankfurt, Germany). The digested PCR–DNA was run on 2 % agarose gel at constant voltage (130 V) for 80 min and visualized by UV. The yeasts were identified on the basis of their restriction patterns using the Spanish Type Culture Collection CECT database (www.yeast-id.org).

Amplification of D1/D2 of 26S rDNA and sequencing

The D1/D2 domain of the large subunit ribosomal RNA gene was amplified with NL1 (5′-GCATATCAATAAGCGGAGGAAAAG-3′) and NL4 (5′GGTCCGTGTTTCAAGACGG-3′) primers [25]. The amplified PCR–DNA of D1/D2 of 22 yeasts which were randomly selected from different RFLP patterns was purified following the manufacturer’s (PeqLab, Erlangen) instructions, and it was sequenced with NL1 primer (5′-GCATATCAATAAGCGGAGGAAAAG-3′). The different yeast strains were identified using the NCBI database (Blastn).

Results

Metabolic spectrum as biomarker for yeast classification and identification

The yeast content of the traditional starter, which is used for Tchoukoutou production, varies from 108 to 109 CFU/g of liquid starter. The colonies of isolated yeasts are creamy-white and yellowish, butyrous in texture, farinose and of smooth appearance. The different yeasts exhibited round, oval and cylindrical shapes at microscopic observation. The fermentation test discriminated the isolated yeasts into different fermentative groups with distinct metabolic profiles. After the classification of the isolated yeasts based on their phenotypic characteristics, two principal clusters were obtained (Fig. 1). The online yeast identification based only on their morphological and physiological characteristics generates ambiguous and inconclusive results. The key procedure described by Back [21] has been used to discriminate S. cerevisiae (lysine agar negative) from the non-Saccharomyces yeast (lysine agar positive). Cluster 2 probably consists of only the non-Saccharomyces yeasts (lysine agar positive) while cluster 1 consists of S. cerevisiae (lysine agar negative) with the exception of the yeast strain BE-227. Most of the lysine and crystal violet negative yeast strains do not ferment melibiose. They therefore have the same fermentative aptitude as S. cerevisiae top-brewing yeast [21]. Twenty-five of the isolated yeasts including S. cerevisiae and non-Saccharomyces are able to use lactic acid as a substrate. Two of the isolated yeasts fermented cellobiose and therefore express β-D-glucosidase (cellobiase). Most (99.58 %) of the isolated yeasts were able to grow at 37 °C (Table 1).

Fig. 1
figure 1

Yeast phenome heat map showing the repartition of the isolated yeasts based on their fermentative aptitude and capacity to grow under stress conditions using the Ward method and Euclidean distance. The color play ranges from green to red via yellow when the reaction is positive, weak or negative

Full size image
Table 1 Physiological characteristics of yeasts isolated from Beninese African sorghum beer starter
Full size table

Proteome as biomarker for yeast discrimination and identification

A hundred and forty-five yeasts with the largest differences in their metabolic profiles were selected and submitted to MALDI-TOF analysis. After this analysis, 92 yeasts were identified, and among them, 69.51 % were identified at species level (score value ≥ 2.3) as P. kudriavzevii and S. cerevisiae, and 30.49 % were identified at genus level as Saccharomyces sp. and Debaryomyces sp. After similarity calculations and multidimensional scaling (MDS) based on the yeast proteome mass spectra, the 145 yeasts were distributed to four clusters (Fig. 2). All the yeasts identified as S. cerevisiae and Saccharomyces sp. as well as 45 non-identified yeasts were grouped in cluster A and those yeasts that were identified as P. kudriavzevii belonged to cluster B. Cluster C consisted of only the yeasts belonging to the Debaryomyces genus and two unidentified yeasts. The out-cluster yeast which does not belong to cluster C was also identified as Debaryomyces. Six of the yeasts that were unidentified because they scored below 1.77 were grouped in cluster D. These yeasts have finally been identified as C. ethanolica on the basis of the combination of the partial domains 1 and 2 of the large subunit ribosomal RNA gene and the restriction patterns of the 5.8S-ITS region. The different mass spectra generated from the different groups obtained after MALDI-TOF MS analysis were different and discriminated the isolated yeasts to genus and species levels (Fig. 3).

Fig. 2
figure 2

Multidimensional scaling (MDS) based on protein patterns showing the repartition of different yeasts isolated from the starter of African sorghum beer. S. cerevisiae yeasts were separated from non-Saccharomyces yeasts. The non-Saccharomyces yeasts were distributed into three different groups belonging to Pichia, Candida and Debaryomyces genus. Saccharomyces cerevisiae is the dominant yeast (A) followed by P. kudriavzevii (B), D. hansenii (C) and C. ethanolica (D)

Full size image
Fig. 3
figure 3

Proteome mass spectrum belonging to each of the four groups of identified yeast: S. cerevisiae (TC-209), D. hansenii (BE-227), P. kudriavzevii (ND-355) and C. ethanolica (SV-256) obtained after matrix-assisted laser desorption/ionization–time-of-flight mass spectrometry (MALDI-TOF MS) using Bruker software, α-cyano-4-hydroxy-cinnamic acid as matrix after the ethanol extraction method

Full size image

Yeast molecular characterization on the basis of ITS1-5.8S-ITS4 region and D1/D2 of large subunit rRNA gene

Fifty isolated yeasts were randomly selected from the different protein patterns (MALDI-TOF MS) and submitted to the restriction fragment length polymorphism (RFLP) analysis. After this RFLP analysis with three different endonucleases (HindIII, HinfI and HaeIII), only HinfI (Hf) and HaeIII (He) digested the ITS1-5.8S-ITS4 region into 2, 3, 4 or 5 fragments according to the strain as shown by the restriction patterns obtained after electrophoresis on agarose gel (Fig. 4). Six different restriction patterns were obtained after RFLP analysis. The restriction profiles 14-Hf (250 and 200 bp) and 14-He (300 and 150 bp) were obtained after the digestion with HinfI and HaeIII, respectively, and correspond to C. ethanolica yeast species. All the yeasts grouped in cluster D, which were obtained after their classification on the basis of protein patterns (MALDI-TOF MS analysis), have the same restriction patterns. The same observation was made with the different yeasts grouped in clusters B and C. In fact, the isolated yeast strains 337 and 375 have the same restriction profile that corresponds to P. kudriavzevii while those of 233, 219 and 221 (Fig. 4b) belong to D. hansenii complex yeast species. The different yeast strains in clusters B, C and D, which were obtained after yeast classification based on their protein patterns, are therefore homogeneous at species level. The RFLP analysis subdivided the S. cerevisiae yeast into different groups corresponding to different genotypes (Fig. 4a and c). The whole strain selected from the three different restriction patterns and belonged to cluster A which includes identified and non-identified yeasts was successfully identified as S. cerevisiae on the basis of D1/D2 domain. The ambiguous results corresponding to C. ethanolica and P. deserticola with a similarity of 100 and 99, respectively, were obtained when the d1 and d2 domain sequence of yeast belonging to cluster D was used. The yeasts grouped in clusters C and D, on the basis of MALDI-TOF analysis, were, respectively, identified as D. hansenii and C. ethanolica by using 5.8S-ITS region (Table 2).

Fig. 4
figure 4

Restriction patterns obtained after digestion of the amplified PCR product of the ITS1-5.8S-ITS4 domain of selected yeasts with the restriction endonuclease HindIII (Hd), HinfI (Hf) and HaeIII (He). All the restriction patterns of a and b belong to S. cerevisiae. C. ethanolica (14), P. kudriavzevii (337, 375) and D. hansenii (233, 219, 221). The 250-bp ladder (L) was used during this analysis

Full size image
Table 2 Restriction patterns of the isolated non-Saccharomyces yeasts obtained with the application of endonucleases HinfI and HaeIII for the digestion of the 5.8S-ITS region and the corresponding yeast species in the Spanish Type Culture Collection (CECT) database
Full size table

Discussion

The yeast characterization on the basis of their morphological and physiological (fermentation and assimilation) traits does not enable to discriminate the non-Saccharomyces. The variability of the fermentative and assimilation profiles of the isolated strains, all belonging to S. cerevisiae yeast species, limits their identification. However, all yeasts which were identified as S. cerevisiae are lysine negative and, contrary to the non-Saccharomyces yeasts (lysine positive), show a distinct fermentative profile (Fig. 1). In contrast to the isolated yeasts’ morphology and physiology, the yeast proteome obtained after MALDI-TOF MS analysis seems to be a good biomarker enabling yeast discrimination and grouping. The restriction patterns of the 5.8S-ITS sequence and comparing the D1/D2 domain of 26 ribosomal LSU sequence show the homogeny inside the different clusters generated on the basis of protein patterns. Despite the accuracy of MALDI-TOF analysis for yeast grouping, the absence of C. ethanolica in the used database hinders its identification. Kurtzman [26], therefore, proposed the expansion of the databases with additional yeast species to achieve complete yeast identification. The MALDI-TOF MS analysis based on microorganism proteome prior to molecular characterization may yet help to accurately group yeast. This rapid yeast grouping may be a good approach for an accurate, high throughput and inexpensive alternative to a large number of yeast characterizations and classifications. However, grouping the D. hansenii complex yeast is difficult and its identification is limited to genus level. The discrimination of yeast belonging to D. hansenii complex such as D. fabryi, D. macquariensis, D. nepalensis, D. propopidis, D. subglobosus, D. tyrocola, D. vietnamensis, D. vindobonensis and solving the limit of the 26S rRNA gene to differentiate these ascomycetous yeasts still remains a challenge. According to the work carried out by Prista et al. [27] and Martinez et al. [28], the Debaryomyces is very sensitive to the presence of some ionic substances such as sodium and potassium, which induce a variation of their proteome. This high susceptibility of D. hansenii complex, causing a small variation of the growth conditions to strongly affect their proteome, will probably preclude obtaining a protein pattern similar to that content in the used library for a reliable Debaryomyces yeast identification. Comparing the D1/D2 of the 26S rDNA sequence does not enhance the result obtained with yeast identification on the basis of the protein patterns. The limits of differentiating yeast belonging to D. hansenii complex due to the high similarity among their D1/D2 domain of the 26S rRNA gene was also observed by Kurtzman et al. [29], Groenewald et al. [30] and Martorell et al. [31]. Sequence analysis of the D1/D2 domain also indicated that C. ethanolica and P. deserticola are closely related, and they were both identified with a similarity of 100 and 99 %, respectively. The 5.8S-ITS region seems to be more discriminative and, contrary to the D1/D2 domain of 26 LSU, allows the differentiation of C. ethanolica from P. deserticola. Indeed, the polymorphic restriction endonuclease HaeIII generated different restriction patterns corresponding to two DNA fragments (110, 310 bp) for C. ethanolica (Table 2) and three DNA sequences (80, 100, 280 bp) for P. deserticola according to the restriction patterns obtained from the used CECT database. The Debaryomyces yeasts were successfully identified to species level as D. hansenii on the basis of the restriction patterns generated from the digestion of the 5.8S-ITS region with HinfI and HaeIII. This result is likely due to the absence of the others Debaryomyces yeasts in the used CECT database. The RFLP analysis of the 5.8S-ITS sequence is limited and does not allow the differentiation of some closely related Debaryomyces yeasts. Martorell et al. [31] shows the limit of the RFLP analysis of the 5.8S-ITS gene to differentiate to species level some Debaryomyces yeast such as D. castellii, D. courdertii, D. nepalensis, D. polymorphus, D. pseudopolymorphus, D. robertsiae, D. udenii and D. vanrifiae. The recent work of Wrent et al. [32] also shows the limit of the RFLP of the 5.8S-ITS region to discriminate the D. hansenii complex such as D. hansenii, D. fabryi and D. subglobosus. After RFLP of the 5.8S-ITS region, the polymorphic endonucleases HaeIII and HinfI, therefore, lead to restriction patterns which discriminated the isolated Candida, Pichia and probably Debaryomyces yeasts to species level and S. cerevisiae to strain level (Fig. 4). Indeed, the yeast strains belonging to S. cerevisiae have been distributed in three different patterns corresponding to three different genotypes of these species. This biodiversity of the S. cerevisiae component in Tchoukoutou starter has been confirmed by their phenotype, characterized by different metabolic profiles (Fig. 1, sub-clusters I) and protein patterns (Fig. 2) even if the D1/D2 sequence analysis does not evidence any high diversity among these isolated Saccharomyces yeasts. Yet the polyphasic approach that integrates yeast phenotype, protein patterns generated by MALDI-TOF MS, and DNA sequence comparison therefore provides reliable information about relationships among yeasts and helps to avoid a mis-identification. Based on this polyphasic approach, it became evident that S. cerevisiae is present in the whole starter and is the dominant or only yeast species (55–100 %) of 11 starters out of the 12 localities where the starters were collected (Fig. 5). This predominance of S. cerevisiae yeast was also observed by Kayode et al. [33], by Djegui et al. [34] and Greppi et al. [35]. The outgrowth of non-Saccharomyces yeasts by S. cerevisiae was also observed during wine production [36]. As observed by Csoma et al. [37], it seems that S. cerevisiae, according to its genome, is more flexible than some other yeasts, which allows this dominant yeast species to more efficiently adapt to the continuously changing environment during fermentation. S. cerevisiae yeast contained in the Beninese traditional sorghum beer starter have different and interesting fermentative aptitudes. Indeed, some specific strains of S. cerevisiae ferment sucrose and therefore express invertase. Djegui et al. [34] indexed S. cerevisiae with invertase gene as a performant strains for the brewing process. In addition to invertase properties, many isolated yeast strains belonging to S. cerevisiae species show a good fermentative profile. They ferment the mains sorghum’s wort sugar and are able to use lactic acid as one of their substrates. These S. cerevisiae yeast species may increase the beer pH during yeast and LAB symbiotic interaction, thereby allowing the growth of LAB that may contribute to further beer maturation and enhance the beer flavor and nutritive value. The co-presence of S. cerevisiae with P. kudriavzevii is observed in the starters collected from Toucountouna (26.67 %), N’dali (45 %) and from Natitingou where P. kudriavzevii is the dominant yeast (61.11 %). P. kudriavzevii has been identified from several West African traditional beverages made from fermented cereals. N’guessan et al. [38] observed its co-presence with S. cerevisiae in Tchapolo, the Ivory Coast traditional sorghum beer. P. kudriavzevii is also the dominant yeast species isolated from Dolo, traditional sorghum beer produced in several localities of Burkina Faso [39]. Annan et al. [40] used P. kudriavzevii in combination with S. cerevisiae to enhance the taste and flavor of Koko, a Nigerian cereal-based African fermented food. C. ethanolica and D. hansenii complex have never been identified from West African sorghum beer. Recently, Visintin et al. [41] observed that C. ethanolica occurs during fermentation of West African cocoa beans and indexed this yeast species to possess several enzymes that may impact the quality of the final product. S. cerevisiae and C. ethanolica were the two different yeast species isolated from the starter collected in Dassa, Savalou and Save. The beer produced in central Benin is also different from that produced in the north of Benin by it organoleptic quality (taste and flavor) and lower alcohol content. The non-Saccharomyces D. hansenii complex yeasts were identified in the starter collected in two localities in northeast Benin. Some strains of D. hansenii may have a great importance and cooperate with S. cerevisiae for the specific aroma compound profile of the Beninese sorghum beer. In fact, the strains BE-219 and BE-225 identified as D. hansenii express cellobiase (β-glucosidase) and may be a potential candidate for beer flavor enhancement. Specific β-glucosidase is able to hydrolyze the aryl-β-D-glucosides from the non-reducing terminus and generates a good precursor from the secondary gluco-conjugate metabolite sequestrated in the cereal grain for beer bioflavouring. D. hansenii is present in many habitats, but has been most isolated from a traditional cheese. Indexed by Romero et al. [42] as a contaminant of intermediate moisture foods, Gori et al. [43] observed the heterogeneity of this yeast species with regard to the cheese flavoring. Indeed, Gori et al. [43] isolated from a cheese, D. hansenii producing high alcohols (2-methyl-1-propanol, 3-methyl-1-butanol, 3-methyl-3-buten-1-ol) and aldehydes (2-methylpropanal, 3-methylbutanal). Even if spontaneous fermentations are ubiquitous in many developed countries in Africa, Asia and South America and constitute the only way of food preservation, the controlled fermentation with the selected performant yeast and LAB from the African traditional starter should be a good alternative to solve African sorghum beer safety problems, increase its shelf life and guarantee the same organoleptic quality for the consumers. The selection of performant yeast and LAB from this traditional starter still remains a challenge for the industrial production of Tchoukoutou as opposed to the existing household production.

Fig. 5
figure 5

Diversity of yeast population of the traditional starter collected in 12 localities of Beninese sorghum beer production

Full size image