INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Fermented maize dough is the starting material from which several West African products are prepared. It contributes to the staple diets of the peoples of the southern and coastal belts of Ghana and other West African countries. The process involves steeping the maize for 24 to 48 h, followed by milling it. The milled maize is reconstituted with water to form a stiff dough which is packed in fermenting troughs and left to ferment spontaneously for 48 to 72 h. The most popular product prepared from this dough is known as kenkey. Previous studies on maize dough fermentation have revealed the complex nature of the microbial interactions that occur, leading to the selection of a defined stable flora which comprises heterofermentative lactobacilli, identified as members of Lactobacillus fermentum or Lactobacillus reuteri, and yeasts, identified as Candida krusei and Saccharomyces cerevisiae (12, 16, 26). Furthermore, acid production during maize dough fermentation together with formation of antimicrobial compounds determines the microbial stability of the products as well as the nonsurvival of enteric pathogens and other microorganisms (23, 26, 33).

It is important that the Lactobacillus species isolated from this food product be classified to both the species and strain levels in order to assess them with technological properties such as production of antimicrobial substances and formation of acids in the maize dough (26). In addition, characterization to the strain level would help in the selection of starter cultures for the production of standardized maize dough. Many of the techniques used for identification of lactobacilli at the subspecies level are variations of restriction fragment length polymorphism analysis (32, 35), restriction endonuclease analysis (REA) (18, 34), ribotyping (29), and PCR, including the randomly amplified polymorphic DNA (RAPD) technique (19, 28, 39). In a previous work (12), the conventional method used in the identification of lactobacilli was not able to distinguish between L. fermentum and L. reuteri due to the similarity in their phenotypes. To be able to monitor cultures during fermentation, it was our objective to establish a method that is able to discriminate and characterize the dominant Lactobacillus species in this fermented product at the strain level. This study describes a method based on the principle of RAPD and the development of two L. fermentum-specific probes which could be used in the hybridization procedure to specifically identify the dominant Lactobacillus species to the subspecies level.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Sampling, isolation, and growth conditions. A total of 172 Lactobacillus species isolates were taken from samples of spontaneously fermented maize dough collected from a major commercial production site in Accra, Ghana, on several occasions. The samples, each weighing 500 to 1,000 g, were taken from raw maize, steep water, and fermented dough at 24, 48, and 72 h. Samples were also taken from surfaces of the steeping tank and corn mill by swabbing, before use, in order to trace the origin of the dominant strains. The surface layers of the maize dough were removed before sampling. Analyses were performed within 2 h of sampling.

Preparation of crude DNA extract for PCR amplification. Cells from 1 ml of the overnight cultures in MRS broth were collected by centrifugation (10,000 × g for 2 min) and washed twice with 1 ml of sterile MilliQ water (MilliQ Plus; Millipore, Molsheim, France) as previously described (19). The cells were resuspended in 100 µl of sterile MilliQ water by vigorous mixing. The cells were disrupted by vigorous shaking with eight glass beads (0.2 mm in diameter) in each tube by using an Eppendorf mixer at full speed (IKA-VIBRAX-VXR; Janke & Kunkel, GmbH, Staufen, Germany) for 1 h at 4°C. Disrupted cells were centrifuged at 10,000 × g for 5 min. Supernatant fluid was used as the source of template DNA for the PCR.

PCR amplification. One microliter of crude DNA extract (supernatant) was used in the PCR, which was carried out in a Thermal Cycler 2400 (Perkin-Elmer, Norwalk, Conn.). Each sample (50 µl in total volume) was amplified in a reaction mixture containing 0.2 mM (each) deoxynucleoside triphosphate (Perkin-Elmer), 2 µM primer 5'-ACGCGCCCT-3' (19), 0.5 µl (50 µg/ml) of Taq polymerase, and 5 µl of 10× PCR buffer (Boehringer Mannheim GmbH, Mannheim, Germany). The reaction mixture was cycled through the following temperature profiles (19): 94°C for 45 s, 30°C for 20 s, and 72°C for 60 s for 4 cycles, followed by 94°C for 5 s, 36°C for 30 s, and 72°C for 30 s for 26 cycles. The PCR was terminated at 75°C for 10 min, and thereafter the mixture was cooled to 4°C. The mixture was stored at 20°C until use.

Gel electrophoresis. Gel electrophoresis was run by applying 20 µl of a sample to submerged horizontal 1.5% type III High-EEO agarose (Sigma) gels (DNA sub cell; Bio-Rad Laboratories, Inc., Hercules, Calif.). Gels were run at 100 V for 2.5 h in TBE electrophoresis buffer (45 mM Tris-base, 89 mM boric acid, 2.5 mM EDTA [pH 8.3]) without cooling. DNA molecular marker VI (0.5 µg; Boehringer Mannheim) was used as a standard. Gels were stained in ethidium bromide (0.3 µg/ml) for 5 min and thereafter washed for 10 min, visualized at 302 nm with a UV transilluminator, and photographed.

Reading of band patterns and numerical analysis. Band patterns on photo negatives were scanned, and data were collected by using the LKB 2400 Gelscan XL program (Pharmacia LKB Biotechnology AB, Uppsala, Sweden) and were then normalized and further processed with Gelcompar 4.0 computer software (Applied Maths, Kortrijk, Belgium). Cluster analysis was generated and evaluated by using the Pearson product moment correlation coefficient (r) and the unweighted pair group algorithm with arithmetic averages.

Cloning and sequencing. RAPD fragments for cloning were purified from 2.0% SeaKem GTG agarose gel (FMC BioProducts, Rockland, Maine) by the QIAquick gel purification kit (Qiagen Ltd., Hilden, Germany) as recommended by the manufacturer. The purified fragments were cloned into the pGEM-T cloning vector (Promega, Madison, Wis.) as recommended by the manufacturer. The ligation mixture was transformed by electroporation into electrocompetent E. coli XL1-Blue MRF' with a gene pulser apparatus as recommended by the supplier (4).

DNA preparation. Pure chromosomal DNA for hybridization experiments was extracted according to the method previously described (17). After the centrifugation (5,000 × g for 15 min) of 10 ml of cultures grown in MRS at 30°C overnight, the pellet was treated with lysozyme and sodium dodecyl sulfate and then phenol-chloroform extracted. DNA preparations were stored in TE buffer (10 mM Tris-HCl, 1 mM EDTA [pH 8.0]) at 5°C.

Plasmid isolation from E. coli. Plasmid DNA was extracted and purified by using columns according to the instructions of Qiagen Ltd.

Southern hybridization. Chromosomal DNA was digested with EcoRI, and restriction fragments were separated on a 0.6% SeaKem GTG agarose gel (FMC BioProducts). The separated fragments were transferred to Hybond N⁺ membrane (Amersham Pharmacia Biotech) by vacuum blotting. The gels were treated for 4 min with 250 mM HCl, followed by 4 min with 1.5 M NaCl-0.5 M NaOH and 3 min with 1 M Tris (pH 7.0)-1.5 M NaCl. The gels were embedded in 20× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate) (30) for 3 h under a vacuum. Probe labeling and hybridization were conducted with the ECL direct nucleic acid labeling and detection kit (Amersham Pharmacia Biotech) as recommended by the manufacturer. NaCl (0.5 M) was added to the hybridization buffer, and washing was performed with primary washing buffer (6 M urea, 5% NaCl, and 2× SSC). Hybridization and washing with primary washing buffer were performed at 42°C, while washing with the secondary wash buffer was performed at room temperature.

Nucleotide sequence accession number. The nucleotide sequences of the upper and lower fragments of LMG 6902^T were deposited in GenBank under accession no. AF090993 and AF090994, respectively.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Microbiological analysis. From the UBA plates, 278 representative isolates from different stages of fermentation (Table 1) were examined; they were all gram positive, nonmotile, microaerophilic, and catalase and cytochrome negative, and they metabolized glucose fermentatively in H & L medium. Among these, 85% were rods (polymorphic), occurring singly or in pairs, and produced gas from glucose in MRS. These were referred to as obligatory heterofermentative lactobacilli, and their numbers during fermentation (24 to 72 h) were in the range of 10⁸ to 10⁹ CFU/g. Good growth was observed for all isolates at 45°C. Most strains did not grow at 15°C, though some growth was observed with some isolates from the raw maize, steep water, steeping tank, and 72-h-fermented dough (results not shown). Fifteen percent of the isolates obtained on the UBA plates were homofermentative lactobacilli with no gas production from glucose in MRS. No further investigation was performed on these isolates. A total of 110 heterofermentative isolates from the present work were selected arbitrarily and further characterized together with other Lactobacillus spp. (Table 1).

RAPD and cluster analyses. A total of 172 isolates (Table 1), which includes 110 Lactobacillus isolates representative of the dominating species from the various steps of maize fermentation, 41 isolates from a previous work (12), and 21 reference strains, were included in the RAPD analysis. All heterofermentative maize dough isolates, including those previously identified as L. fermentum or L. reuteri (12), and the L. fermentum type (LMG 6902^T) and reference strains exhibited two amplified DNA fragments with sizes of 695 and 773 bp (Fig. 1). These fragments were not apparent in the other Lactobacillus species tested except L. reuteri LMG 13045 (results not shown). Cluster analysis was performed on all 172 isolates (Table 2). The subclustering positions of the 172 isolates are summarized in Table 2, and representative isolates are shown in Fig. 2. Two main clusters were observed: cluster 1 included all L. fermentum strains, and cluster 2 included the remaining strains of Lactobacillus examined (Table 2). The two clusters emerged at a similarity level of about <50%. Within cluster 1, four distinct subclusters were defined at a similarity level of about 75%, based on their RAPD profiles and considering only the strong and reproducible band patterns. Strains included in subcluster 1A had a reproducible additional band at approximately 1,766 bp. This cluster contains strains from all stages of fermentation but not from the steeping tank and fresh dough. Strains in subcluster 1B had two reproducible additional bands at about 1,766 and about 350 bp. This group, unlike subcluster 1A, included strains only from the steep maize, fresh dough (0 h), and 48-h-fermented dough. Strains included in subcluster 1C had no additional band. This subcluster comprised strains from all steps of the fermentation but not from the steep water. Subcluster 1D was defined by a band at approximately 350 bp, and it also included strains from all steps of the fermentation (Fig. 2 and Table 2). Only two of the L. fermentum reference strains, LMG 8900 and LMG 8902, and L. reuteri LMG 13045 clustered with maize isolates in subcluster 1C. The remaining L. fermentum reference strains LMG 8899, LMG 11441, LMG 8154, and LMG 8896, including the type strain LMG 6902, clustered together outside the main clustering of maize dough strains but within the main cluster 1. 

View larger version (99K): [in this window] [in a new window]   FIG. 1.   (A) Examples of RAPD profiles of selected reference strains of L. fermentum showing two distinct bands at 773 and 695 bp. Lanes 2 to 7 and 10, LMG 11441, LMG 8900, LMG 8896, LMG 8154, LMG 8902, LMG 8899, and LMG 6902T, respectively; lane 14, maize dough isolate (A9-18) (note the absence of these bands in other Lactobacillus spp.); lanes 9 and 11 to 13, L. reuteri LMG 9213T, L. leichmannii ATCC 7830, L. plantarum ATCC 8014, and L. casei ATCC 7467, respectively; lanes 1, 8, and 15, DNA marker VI (Boehringer Mannheim) (molecular sizes are indicated). Arrows a, b, c, and d indicate the 1,766-, 773-, 695-, and 350-bp fragments, respectively. (B) RAPD profiles of selected L. fermentum strains from maize dough showing the two distinct bands at 773 and 695 bp. Lanes 2 to 6, 10, and 11, A7-15, A7-11, A6-7, A5-11, A4-25, B6-6, and B6-4, respectively (note the absence of these bands from other species); lanes 7, 9, and 12 to 14, L. buchneri A1-24, L. plantarum B11-1, L. plantarum B2-1, and unknown, respectively; lanes 1, 8, and 15, DNA marker VI (Boehringer Mannheim). Arrows a, b, and c indicate the 1766-, 773-, and 695-bp fragments, respectively.

View larger version (71K): [in this window] [in a new window]   FIG. 2.   Dendrogram illustrating an example of the clustering of representative L. fermentum isolates from maize dough and other Lactobacillus species based on their RAPD profiles, which were evaluated by using the Pearson product moment correlation coefficient (r) and the unweighted pair group algorithm with arithmetic averages.

Characterization of specific RAPD fragments. Restriction digests of two purified dominant RAPD fragments from L. fermentum A4-25 and LMG 6902^T indicated that the upper and lower fragments were each composed of one main DNA fragment (data not shown). In order to evaluate if the two dominant RAPD fragments of L. fermentum strains were conserved, we decided to clone them from L. fermentum A4-25 and LMG 6902^T, respectively. The plasmids with upper fragments, pLBUF (A4-25) and pT3UF (LMG 6902^T), had inserts of 773 bp, while the plasmids with lower fragments, pLBLF (A4-25) and pT3LF (LMG 6902^T), had inserts of 695 bp. The inserts of two to three clones of each of the four fragments were DNA sequenced on both strands. The DNA sequence analysis revealed only minor strain-specific differences between the sequences of the upper and lower fragments, as shown in Table 3. The G+C content of the sequenced fragments was between 53.7 and 54.8%, which is close to the 52 to 54% G+C content reported for L. fermentum (13).

View larger version (54K): [in this window] [in a new window]   FIG. 3.   Comparison of the predicted amino acid sequence of the upper fragments from L. fermentum A4-25 (LBUF_LF) and LMG 6902T (T3UF_LF) with half the amino acid sequence of D-hydroxyisocaproate dehydrogenase of L. delbrueckii subsp. bulgaricus (DHDH_LD) (accession no. S48145) (2) and L. casei (DHDH_LC) (accession no. P17584) (22); with D-lactate dehydrogenase from Pediococcus acidilactici (DLDH_PA) (accession no. X70925), L. plantarum (DLDH_LP) (accession no. P26298), and L. delbrueckii subsp. bulgaricus (DLDH_LD) (accession no. X60220); and with D-hydroxy acid dehydrogenases from Staphylococcus aureus (DDH_SA) (accession no. U31175), Enterococcus faecium (DDH_EF) (accession no. O05709), and Enterococcus faecalis (DDH_EFL) (accession no. U35369). Amino acids identical to those of L. fermentum LMG 6902T are blue on a yellow background. Amino acids identical in all sequences are white on a green background. The aspartic acid in the first red bar is the putative amino acid that distinguishes NADH from NADPH (2). The arginine in the second red bar constitutes the putative substrate binding site (10, 21), and the histidine in the third red bar is the putative proton donor (10, 36).

Hybridization studies and development of DNA probe. The specificity of the conserved RAPD fragments was tested by using Southern hybridization. Labeled DNA probes from the conserved 773-bp (pLBUF) and 695-bp (pLBLF) fragments from L. fermentum were used. L. fermentum strains tested were selected from the three major subclusters (1A, C, and D). Results are shown in Fig. 4. The two probes hybridized strongly to all 15 L. fermentum strains (Fig. 4A and B, lanes 1 to 13, 17, and 18) but not to other Lactobacillus species (Fig. 4A and B, lanes 14 to 16; Fig. 4C and D, lanes 1 to 5, 7, 9 to 11, and 13 to 17). Hybridization signals were observed for L. reuteri LMG 13045 (Fig. 4C and D, lane 12). This strain also had the same RAPD pattern as L. fermentum strains (Fig. 2). Two very faint hybridization signals were observed for L. reuteri DRO 2010 and L. plantarum 1-6A (Fig. 4C and D, lanes 6 and 8) on overexposed film. Six groups could be identified for the L. fermentum maize isolates analyzed, based on the sizes of the EcoRI fragments hybridizing with the upper (pLBUF) and lower (pLBLF) fragments, respectively, for the following. For group 1 (strain G12), the probes hybridized to 11.5- and 6.0-kbp fragments (Fig. 4A and B, lanes 2). For group 2 (strains H4, A7, and B12), the probes hybridized to 11.5- and 6.5-kbp fragments (Fig. 4A and B, lanes 4, 6, and 9). For group 3 (strains D17, C2, and A7-11), the probes hybridized to an approximately 20-kbp fragment and a 11.5-kbp fragment (Fig. 4A and B, lanes 1, 7, and 18). For group 4 (strains H11 and C14), the probes hybridized to an approximately 6.3-kbp fragment and an approximately 13-kbp fragment (Fig. 4A and B, lanes 3 and 12). For group 5 (strains A31, G20, A28, and A4-25), the probes hybridized to 6.3- and 11.5-kbp fragments (Fig. 4A and B, lanes 5, 8, 11, and 17). Finally, for group 6 (strain H17), the probes hybridized to a 6.5-kbp fragment and an approximately 13-kbp fragment (Fig. 4A and B, lane 10). The type strain LMG 6902 had hybridization signals differing from those of the maize isolates, e.g., approximately 16- and 7.0-kbp fragments (Fig. 4A and B, lanes 13).

View larger version (25K): [in this window] [in a new window]   FIG. 4.   (A and B) Hybridization signals obtained with pLBUF DNA (A) and pLBLF DNA (B) probes to a membrane with EcoRI-digested and separated chromosomal DNA. Lanes 1 to 13 and 17 to 18, L. fermentum strains D17, G12, H11, H5, A31, A7, C2, G20, B12, H17, A28, C14, LMG 6902T, A4-25, and A7-11, respectively; lane 14, L. brevis LMG 6906T; lane 15, L. confusus LMG 6898T; and lane 16, L. reuteri LMG 6892T. (C and D) Hybridization signals obtained with pLBUF DNA (C) and pLBLF DNA (D) to membrane with EcoRI-digested and separated chromosomal DNA. Lanes 1 to 4 and 8 to 11, DNA from L. plantarum strains C11, L11, L30, LMG 6907T, 1-6A, 2-29A, 1-6B, and 1-9B, respectively; lanes 5 to 7 and 12 to 17, L. reuteri strains DRO 2000, DRO 2010, DSM 20016, LMG 13045, LMG 13046, LMG 13088, LMG 13089, LMG 13090, and LMG 13091, respectively.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The present investigation supports previous findings (12) that heterofermentative Lactobacillus spp. make up the numerically dominant lactic acid bacteria in Ghanaian fermented maize dough. The fermentation of this indigenous product is spontaneous and requires no inoculum. The presence of the bacteria on raw maize, the steeping tank, and the corn mill may indicate the likely sources of the bacteria and also the possible stages in the production line where natural inoculation could occur during the processing of the fermented product. Due to the spontaneous nature of fermentation and the standard level of production in Ghana, maize dough often results in a product of variable quality. In order to standardize this fermentation process and to be able to produce a standard-quality product, the production of starter cultures is being suggested. To produce a starter culture, however, a precise identification procedure is required. Previous studies (12, 26) could not differentiate between L. fermentum and L. reuteri by phenotyping. We proceeded to use DNA-based methods to achieve this objective. To further characterize the heterofermentative and previously identified strains from fermented maize dough, a RAPD method was applied. The method was able to differentiate between L. fermentum and L. reuteri by the generation of several species-specific bands. The RAPD method reproducibly generated two strong species-specific bands in all heterofermentative isolates obtained in this work, as well as in previously identified heterofermentative isolates from maize dough and from type and reference strains of L. fermentum. The absence of these dominant fragments in L. reuteri strains, as well as in other species of lactobacilli tested (Table 1), confirmed L. fermentum and not L. reuteri as the dominant lactic acid bacterium in the fermentation of maize dough. The dominant role of L. fermentum in maize dough and similar products in the West African region has been indicated (5, 14). RAPD analysis also generated additional bands which allowed for differentiation within the main L. fermentum group. Based on the cluster analysis results, RAPD patterns revealed four major subgroups within the main L. fermentum group. Strains from different stages of fermentation were recorded in each subcluster, indicating that several strains were involved in the maize fermentation. The presence of strains from the three subgroups (1A, 1C, and 1D) on the raw maize kernels suggested the presence of different strains at the onset of fermentation, and an additional contribution to subgroups 1C and 1D was from the steeping tank. Strains from all four subgroups were detected throughout the fermentation (24, 48, and 72 h), and there appeared to be no selection for any particular subgroup. The fact that strains from different stages of the fermentation clustered together suggests that they remain dominant throughout the fermentation. There was also evidence that significant changes in the population of L. fermentum had not occurred in over 4 years, i.e., all the previous isolates (12) were found in subclusters 1A, 1C, and 1D and not outside the main cluster 1. This result may be due to the fact that these experiments were conducted at the same production site. It is certain that long-term processing of this product at this site has resulted in the residence of these organisms in the environment and processing equipment (steep tank and corn mill). Hence, growth of the organisms is likely to be stimulated in the presence of the substrate (maize).

The RAPD method used in this investigation was able to discriminate at the species level and below, and the discriminating power of this method has been confirmed by others (8, 38). The method is increasingly used due to its ease of operation, although constant references are made to the fact that reproducibility between laboratories becomes almost impossible due to the method's dependence on the type of Taq polymerase and thermal cycler used (24, 37). On the other hand, the method appears to have some merit when it comes to generating genetic markers for linkage mapping (11) and creating species-specific probes (9, 28) when no sequence data are available for the genome in question. In the present work, the method was found to be reliable and reproducible except for variations that occurred with bands of low intensity. In the present investigation, the shelf life of the PCR reagents was crucial. All the experiments were repeated several times, independently, and the classification of patterns of amplification was based only on strong and reproducible bands.

To investigate the specificities of the two prominent conserved RAPD bands in L. fermentum, we developed probes based on the two amplified fragments from two L. fermentum strains (LMG 6902^T and A4-25). The two amplified RAPD fragments were cloned, sequenced, and used as probes in hybridization experiments for the detection of L. fermentum. Based on the sequencing results, differences in the DNA sequences of the upper and lower fragments between both strains (L. fermentum LMG 6902^T and A4-25) were minimal, only 5 and 6 bp, respectively, indicating a conserved region within the genome of these strains. The sequence from the upper fragment revealed homology to the C-terminal part of D-hydroxyisocaproate dehydrogenases and D-lactate dehydrogenases. It contained the conserved amino acids usually found in this region of 2-hydroxy acid dehydrogenases (Fig. 3). The sequence of the lower fragment showed homology to two putative proteins from B. subtilis with an unknown function. It contained two phosphate-binding loop motifs, indicating that it might be an ATP-GTP binding protein. Except for L. reuteri LMG 13045, probes developed from these conserved fragments hybridized strongly to DNA derived from all L. fermentum strains tested and not to the other Lactobacillus spp., thus indicating the specificity of these probes. Two other strains, L. reuteri DRO 2010 and L. plantarum 1-6A, gave weak signals when the film was overexposed; however, these signals could be clearly differentiated from the much stronger hybridization signals of strains belonging to L. fermentum. In addition to a positive hybridization with the probes, L. reuteri LMG 13045 also had the typical RAPD profile of L. fermentum strains. Information from LMG stated that the strain was originally classified as L. fermentum and later reclassified as L. reuteri based on protein profiles. This emphasizes existing doubts as to the accuracy of the criteria used to reclassify this strain. To support this view, some workers (38) reported that protein profiles sometimes fail to discriminate between closely related species, as with L. plantarum and Lactobacillus pentosus; however, the RAPD method was able to separate them independently into two separate species. Nevertheless, the identification of L. reuteri LMG 13045 needs further investigation. So far, REA is the only DNA-based method that has been used to distinguish between L. fermentum and L. reuteri (34). This method, although powerful, is laborious, and results are difficult to analyze due to the numerous REA bands that are produced compared to the fewer bands in RAPD analysis. Nevertheless, the specificity of the probes is sufficient to clearly identify L. fermentum strains. The differences in the positions and sizes of the hybridization signals observed in L. fermentum isolates further suggest the diversity among the L. fermentum strains and within each subgroup.

We conclude that several strains of L. fermentum are involved in the fermentation process, and it appears that there is no selection for any particular subspecies variation. For effective selection, monitoring, and control of starter cultures when they are developed, the RAPD method appears to be a useful and fast technique that can be applied. However, Southern hybridization could be advantageous for confirmation of starter cultures and more precise groupings. Through our findings we have located two conserved regions within the genome of L. fermentum. Due to the specificity of the probes to L. fermentum, these conserved sequences may also make it possible to design long primers to be used in PCR-based detection of L. fermentum.