The LAB isolates used in this study (Table 1 (18, 25)) were taken from the culture collections of the SEAFAST Center, IPB University, Bogor, Indonesia. These isolates exhibited varying growth abilities in folate-free media (18). Lactiplantibacillus plantarum WCFS1 was used as a positive control for folate gene detection because all folate biosynthesis genes are present in its genome (24–26). All LAB isolates were stored in a mixture of de Man, Rogosa and Sharpe broth (MRSB; CM0359, Oxoid Ltd., Basingstoke, UK) and 20 % glycerol at −20 °C and revived in MRSB before use.

Each pure culture was grown for 18 h in MRSB supplemented with 10 % glycine. Genomic DNA was extracted from bacterial cell pellets using the Wizard® Genomic DNA Purification Kit (A1120; Promega Corporation, Madison, WI, USA), following the manufacturer's instructions (27). The purity and concentration of the extracted DNA were determined using a Nanodrop ND-1000 spectrophotometer (Thermo Fisher Scientific Inc., Waltham, MA, USA). The DNA samples were then stored at −20 °C.

The presence of eight genes involved in the folate biosynthesis pathway was detected in various LAB species by polymerase chain reaction (PCR) amplification. The primers for the detection and sequencing of each gene are shown in Table 2 (24). Six of these genes were designed using the online tool NCBI Primer-BLAST (the Basic Local Alignment Search Tool) (28) and gene sequences of Lactiplantibacillus plantarum WCFS1 obtained from the Kyoto Encyclopedia of Genes and Genomes (KEGG) (26) and GenBank (29) databases. All primers were manufactured by Integrated DNA Technologies Pte. Ltd. (Coralville, IA, USA).

PCR amplifications were performed in 20 µL volumes with 10 µL Promega Go Taq Green Master Mix 1× (M7122; Promega Corporation), containing Taq DNA polymerase, dNTPs, MgCl₂ and reaction buffers, 1 µL of 0.5 µM of each forward and reverse primer, 1 µL DNA template (>150 ng) and 7 µL nuclease-free water (NFW P1193; Promega Corporation). The amplifications were carried out using an Applied Biosystems 2720 Thermal Cycler (Applied Biosystems, Foster City, CA, USA) with the following cycling conditions: 1 cycle at 95 °C for 2 min; 30–35 cycles at 95 °C for 30 s, the annealing temperature depending on the melting temperature of each primer and tested at 50–60 °C for 1 min, and at 72 °C for 30 s; and 1 cycle at 72 °C for 5 min. The positive and negative controls were also subjected to amplification. The positive control was DNA from the reference strain (L. plantarum WCFS1), while the negative control was NFW without DNA containing the target genes. The PCR products were then separated by gel electrophoresis (Mini-Sub Cell GT Horizontal Electrophoresis System; BioRad Laboratories Inc., Hercules, CA, USA) using 2 % agarose gel in 1× Tris-acetate-EDTA (TAE) buffer and visualized using ethidium bromide staining.

The detected folate biosynthetic genes in six folate-producing isolates of different species, represented by Lactiplantibacillus plantarum 4C261, Lacticaseibacillus rhamnosus R23, Limosilactobacillus fermentum JK13 and BG7, Pediococcus acidilactici NG64, and Leuconostoc mesenteroides S2SR08, were sequenced by 1st Base Sequencing (Selangor, Malaysia) using the Sanger method (Sanger dideoxy sequencing) (30) after purification. A single sample of purified DNA for each folate gene was needed as a template for sequencing. The PCR products with more than one band on agarose gel electrophoresis were first cut out of the gel and purified using a DNA gel extraction kit (GeneJET Gel Extraction Kit; Thermo Fisher Scientific Inc.). Samples with a single DNA band of the expected size on agarose gel electrophoresis were purified further before sequencing. The PCR cleanup step was then performed by an ultrafiltration method using Centricon-100 micro-concentrator columns (Applied Biosystems) to remove unincorporated primers and dNTPs that can interfere with the sequencing results. DNA quality and quantity were then determined by agarose gel electrophoresis and spectrophotometry. After DNA template preparation, cycle sequencing was performed using the BigDye Terminator v3.1 cycle sequencing kit (Applied Biosystems) and run in a thermal cycler (GeneAmp PCR System 9700; Applied Biosystems). The excess dye terminator was removed by spin column purification with Centri-Sep spin columns (Applied Biosystems) prior to analysis on an ABI PRISM 3730xl genetic analyzer (Applied Biosystems). The primers used for sequencing were those listed in Table 2 (the same primers for gene detection). The sequencing output was then analyzed using Sequence Scanner Software v. 2.0 (Applied Biosystems) (31).

The sequenced nucleotides were processed using MEGA7: Molecular Evolutionary Genetic Analysis v. 7.0 (32). The obtained consensus DNA sequences (contigs) were used as the query sequences to perform homology searches using the NCBI BLAST algorithm (28). The 'blastn' program was used to find areas of local similarity between query and database nucleotides. The identified query sequences were then stored in the GenBank database with the accession numbers shown in Table 3 (29).

Due to the inadequacy of the folate biosynthesis gene database in NCBI GenBank (29), the KEGG (26) was used as a primary reference for nucleotide sequence database for each folate biosynthetic gene. The MEGA v. 7.0 software (32) was used to align folate-related gene sequences from six isolates and reference data by the ClustalW program (incorporated in the MEGA software). The same software was also used to create phylogenetic trees using the neighbor-joining method and Jukes and Cantor models for determining evolutionary distance values with a 1000 replicate bootstrap test.

The inoculum of all LAB isolates was prepared according to a previously described method (18). Bacterial cell pellets grown for 24 h in MRSB were washed twice by centrifugation (refrigerated centrifuge LRF-B20; Labtron Equipment Ltd., Camberley, UK) at 10 000×g for 5 min at 4 °C and finally resuspended in sterile saline solution (0.85 % m/V NaCl; Merck KGaA, Darmstadt, Germany). The bacterial suspensions had cell densities of around 9–10 log colony-forming units (CFU)/mL, which were diluted to 5–6 log CFU/mL before use as inoculum. A total of 2 % inoculum was then grown in folate-free medium (folic acid casei medium, FACM; HiMedia Laboratories, Mumbai, India) at 37 °C for 24 h and subcultured twice in the same medium under the same growth conditions. The extracellular folate was extracted by centrifugation (refrigerated centrifuge LRF-B20; Labtron Equipment Ltd.) at 10 000×g for 5 min, followed by supernatant filtration using a 0.2-μm nylon filter membrane (ANPEL Laboratory Technologies Inc., Shanghai, PR China). All samples were then stored at –20 °C until further analysis.

Extracellular folate was analyzed using an Agilent 1260 Infinity HPLC System with a diode array detector (DAD; Agilent Technologies Inc., Santa Clara, CA, USA) and a ZORBAX Eclipse XDB-C18 chromatography column (15 cm×4.6 mm, 5 μm; Agilent Technologies, Inc.) at λ=282 nm. The mobile phase was freshly prepared and consisted of water (HPLC grade; LiChrosolv®, Merck KGaA) with glacial acetic acid (0.66 %; EMSURE®, Merck KGaA) and methanol (pure HPLC grade; LiChrosolv®, Merck KGaA), with a ratio of V(water):V(methanol)=70:30 (33). The flow rate was 0.8 mL/min. Folic acid standard was obtained from R-Biopharm (provided in the Vitafast folic acid test kit; Pfungstadt, Germany) and used without further purification.

The data were analyzed using SPSS v. 20.0 (34), with the statistical significance level set at 95 % (α<0.05). A one-way ANOVA was used to compare the differences in the extracellular folate productivity of the isolates in folate-free culture medium (FACM).

Eight folate biosynthetic genes (folE, folQ, folB and folK in the DHPPP formation pathway, and folP, folC₁, folA and folC₂ in the THF-polyglutamate formation pathway) were successfully amplified using gene-specific primers, with the predicted size of the PCR amplicon for each gene (Fig. 1). As a positive control, a single band for each of the eight genes was also detected in Lactobacillus plantarum WCFS1. No bands were detected in the negative control (data not shown).

However, despite designing specific primers and performing amplifications at various annealing temperatures, non-specific detections, seen as a double band or multiple bands on gel electrophoresis, were observed. In seven of the 11 folate-producing isolates, i.e. Lactobacillus rhamnosus R23 and R15, L. fermentum JK13, L. plantarum 4C261, R12 and JK16, and Pediococcus acidilactici NG64 (18), all folate biosynthetic genes were specifically detected. In the other folate-producing isolates, there were non-specific detections of one gene, i.e. L. fermentum BK27 (folB), L. fermentum BG7 (folB), Leuconostoc mesenteroides S2SR08 (folC₁) and L. rhamnosus BD2 (folP). Non-specific detection of genes was previously reported by Saubade et al. (35), where the folP gene was detected in several pearl-millet-based porridge samples by more than one band on gel electrophoresis. The report also suggested that both specific and non-specific detections were thought to indicate the presence of a gene.

Two isolates, JK6 and BG8, are known folate non-producers (18) and did not have a complete set of folate biosynthesis genes, possibly leading to their inability to produce folate. The JK6 isolate lacked the folQ gene, while the BG8 isolate lacked folQ, folP and folA, as indicated by the absence of the corresponding bands after gel electrophoresis (Fig. 1b, Fig. 1e and Fig. 1g). However, the two non-folate-producing isolates still have folK, one of the signature genes for folate biosynthesis, and even the JK6 isolate has both signature genes, folK and folP. This finding is supported by Greppi et al. (9), who reported that some LAB isolates could not synthesize folate despite having the signature genes folP and folK. Other folate biosynthetic genes probably play an essential role in determining the LAB folate production capacity.

Tetrahydrofolate, whose production is regulated by folP, folC₁, folA and folC₂, acts as a cofactor in one-carbon metabolic reactions in various pathways, such as the biosynthesis of purines, thymidine, glycine, methionine, pantothenate and formyl-methionyl tRNA (fMet-tRNA), which is required for the initiation of protein synthesis (23, 36–39). These folate-dependent metabolites, except fMet-tRNA, can be provided in bacterial growth media. Thus, THF is essential for the production of fMet-tRNA, which is needed for the growth of folate-requiring bacteria. In the final stage of the THF biosynthetic pathway, the reduction of dihydrofolate to THF is catalyzed by dihydrofolate reductase (DHFR), encoded by the folA gene. In addition, when bacteria require a folate supply from the environment, the conversion of folic acid to dihydrofolate in the folate salvage pathway also involves the activity of the DHFR enzyme (23). For this reason, all folate-dependent bacteria need a minimum amount of DHFR activity to synthesize fMet-tRNA. Thus, DHFR is expected to be found in both folate producers and non-producers (23, 40).

However, not all LAB isolates in our investigation had the folA gene. The BG8 isolate, a non-producer of folate, lacked this gene. Levin et al. (41) also stated that although THF is an essential cofactor for all bacteria, the gene encoding DHFR cannot be found in many bacteria. The unavailability of the folA gene in these bacteria may be due to the presence of another type of DHFR enzyme encoded by another gene. Work by de Crécy-Lagard et al. (23) showed that some bacteria might have different types of DHFR, such as DHFR1 (encoded by folM), belonging to the short-chain dehydrogenase/reductase family, or DHFR2, a flavin-dependent dihydropteroate reductase fused with DHFR. Hence, the BG8 isolate is predicted to also have the folM gene or the gene encoding DHFR2, as a homologous gene of folA.

In the folate biosynthetic pathway, the formation of THF-polyglutamate is crucial for folate retention in bacterial cells and is required for most folate-dependent enzymes which have a higher affinity for polyglutamate folate (6, 42, 43). The conversion of THF monoglutamate into the polyglutamate form is catalyzed by folylpolyglutamate synthase (FPGS), encoded by the folC₂ gene. Therefore, in addition to folA, folC₂ should also be found in all folate-dependent bacteria. Furthermore, FPGS is essential for folate-consuming bacteria, as they can only sequester monoglutamate folate (1-3 glutamate residues) via the salvage route due to the lack of an enzyme to break down folate polyglutamate (γ-glutamyl hydrolase) (20, 44, 45). However, in some bacteria, the folC₂ gene can be found in the fusion gene of folC, encoding the bifunctional enzymes DHFS and FPGS. Non-folate-producing bacteria lacking either the folC₂ or folC genes may harbour a gene encoding a novel type of FPGS. According to de Crecy-Lagard et al. (23), the unavailability of the folC gene in Mycoplasma species relying on the salvage pathway may be due to the presence of another gene encoding a different FPGS. In this study, folC₂ was detected in all LAB isolates (producers and non-producers of folate).

Since all folate-dependent bacteria have the genes folA and folC₂ (or folC), or their homologues (23, 40), folate biosynthetic genes defining the potential ability of LAB to synthesize folate may include genes folE, folQ, folB and folK, involved in the DHPPP biosynthetic pathway, and the folP gene, involved in THF-polyglutamate biosynthesis. Rossi et al. (12) also reported that the genes folC (folC₁/folC₂) and folA, or their homologues, were found in all studied LAB species, whereas other folate biosynthetic genes were detected only in a few species.

In this study, the JK6 and BG8 isolates lacked the gene folQ, which encodes dihydroneopterin triphosphate pyrophosphohydrolase. In some folate-producing bacteria (such as Bacillus subtilis, Escherichia coli and Streptococcus thermophilus), the absence of the folQ gene may be due to the presence of other homologous genes that have not been identified (25). However, as JK6 and BG8 isolates could not grow in a folate-free medium (FACM), the missing folQ gene in both isolates may indicate that the gene folQ can be a limiting factor in their ability to synthesize folate. Moreover, Liu et al. (46) have reported that in the DHPPP biosynthetic pathway, the folQ gene had the highest expression level, demonstrating a better folate synthesis ability in L. plantarum strain 4_3 used in their study. Although further research of gene expression level analysis is also required in this study, the present findings suggest that the folQ gene may play a crucial role in determining the potential ability of LAB to synthesize folate.

Here, the detection of folate biosynthetic genes did not include the detection of genes encoding PABA biosynthetic enzymes (pabA and pabB/pabC) due to the lack of these genes in almost all lactobacilli, as reported by Rossi et al. (12). Thus, lactobacilli are generally unable to produce folate without PABA supplementation in the medium (9, 12). The FACM used in this study has neither folate nor glutamic acid but it still contains 2 mg/L of PABA (based on the technical datasheet). Despite the presence of PABA in the FACM, the JK6 and BG8 isolates were unable to grow in this medium, indicating that both could not utilize PABA as a precursor in the medium to carry out folate biosynthesis in their cells. Hence, the availability of PABA biosynthetic genes can be neglected in this case.

The sequences of eight folate biosynthetic genes (folE, folQ, folB, folK, folP, folC₁, folA and folC₂) from six folate-producing LAB isolates (R23, 4C261, JK13, BG7, S2SR08 and NG64) (using the same primer for gene detection) yielded different contig sizes for each isolate (Table S1). Nucleotide BLAST search revealed that all the folate genes are homologous with the LAB genome (instead of specific genes for folate biosynthesis) in the NCBI database with 100 % nucleotide sequence identity (Table S2). Hence, the references of nucleotide sequences for each folate biosynthetic gene from KEGG database were used to construct the phylogenetic trees shown in Fig. 2 and Fig. 3. The phylogenetic tree analyses showed that the six folate genes (folE, folQ, folB, folA, folC₁ and folC₂) found in folate-producing isolates are from the monophyletic group of LAB species in the database, including the positive control, L. plantarum WCFS1 (Figs. 2a-2c and Figs. 3b-3d). For the two other genes, folK and folP, only isolate R23 has both genes from the monophyletic group of LAB species in the database, while the other five folate-producing isolates have the genes from the polyphyletic group (Fig. 2d and Fig. 3a).

L. plantarum WCFS1 was selected as a reference strain to compare the similarity of the folate genes from the six isolates because it possesses a complete set of folate biosynthetic genes and has been well studied for its capacity to produce folate (9, 24–26). The comparison of the number of conserved sites (C) and variable sites (V) of eight folate biosynthetic genes among the six folate-producing LAB isolates and the reference strain (L. plantarum WCFS1) is given in Table S3. A higher percentage of conserved sites was found in six folate genes that are monophyletic with L. plantarum WCFS1, ranging from 66 to 86 % for folE, 47–96 % for folQ, 96–98 % for folB, 95 % for folC₁, 66–97 % for folA, and 92–96 % for folC₂. Only 0–2 % of sites were variable sites and the rest were gaps. Meanwhile, the two folate genes (folK and folP) of five isolates that are polyphyletic with L. plantarum WCFS1 showed fewer conserved sites (21–30 % for folK and 20–23 % for folP) than the variable sites (33–53 % for folK and 39–54 % for folP). As an exception, R23 had a higher proportion of conserved sites for both genes, as it is monophyletic with WCFS1.

Six folate-producing LAB isolates have the genetic capacity for de novo folate biosynthesis. The R23 isolate had all homologous folate genes that were the same as the positive control, L. plantarum WCFS1. The nucleotide sequences of the folK and folP genes of the five isolates showed remarkable differences from those of the positive control. The higher variable sites of folK and folP in five LAB isolates may indicate that both genes are not essential in the folate biosynthesis pathway. Jordan et al. (47) stated that essential genes are more conserved than nonessential genes in bacteria. The folK and folP genes of five isolates may contribute at different levels to de novo folate biosynthesis. However, more research is required to determine their correlation with folate production ability. The selection of folQ as a marker gene for folate biosynthesis in this study may also indicate that the folQ gene is essential in the folate biosynthesis pathway.

Based on the phylogenetic tree in Fig. 2b, the gene folQ, which encodes the enzyme that catalyzes the conversion of dihydroneopterin triphosphate to dihydroneopterin monophosphate, is known to produce two isoforms, XTP/dITP diphosphatase (EC 3.6.1.66) and dihydroneopterin triphosphate diphosphatase (EC 3.6.1.67), that have the same function. Several LAB species from the KEGG database, such as Latilactobacillus sakei, L. kefiri and Lactococcus lactis, have the folQ gene encoding dihydroneopterin triphosphate diphosphatase (EC 3.6.1.67), while L. plantarum species has the folQ gene, which encodes for the XTP/dITP diphosphatase. These two types of enzymes were divided into two different clades (Fig. 2b). The folQ genes of six folate-producing LAB isolates (R23, 4C261, JK13, BG7, S2SR08 and NG64) in this study were in the same cluster as L. plantarum, indicating that the enzyme encoded by the folQ gene of the six isolates was probably the XTP/dITP diphosphatase.

The concentration of folate produced by eleven LAB isolates (4C261, R12, R23, R15, BD2, JK13, JK16, BK27, BG7, S2SR08 and NG64) ranged from 10.37 to 31.10 μg/mL (Fig. 4a). The variability in the concentration of folate produced by these isolates was related to the availability of all eight genes involved in synthesizing folate (Fig. 4b). However, two isolates, JK6 and BG8, did not secrete extracellular folate (Fig. 4a) and did not have all the genes for folate biosynthesis (Fig. 4b). Five isolates (R23, JK13, BD2, R15 and NG64) had higher extracellular folate productivity, ranging from 18.40 to 31.10 μg/mL, than that of the positive control, L. plantarum WCFS1. The isolate L. rhamnosus R23 was the highest folate producer, producing 207 % more folate than WCFS1.

The folate concentrations measured in the isolates in this study were higher than those reported by Kodi et al. (33), who studied the LCF10 LAB strain isolated from fermented milk. Using the same analytical method employed here, they found that LCF10 produced 10.4 μg/mL folate in FACM after 18 h of incubation. Wu et al. (48) found higher folate production concentrations in L. casei, L. acidophilus and L. plantarum, which produced 45.41, 42.78 and 63.23 μg/mL of folate in yogurt products, respectively. Nevertheless, the highest folate concentration (31.10 μg/mL) produced by isolate R23 in this study is equivalent to ~8 % of the recommended daily folate intake, which is 400 μg per day for the average adult (49). Supplementation of food products with the R23 isolate can potentially be used as a good source of natural folate.