Phylogenetic analysis of Leuconostoc and Lactobacillus species isolated from sugarcane processing streams

Abstract High levels of gums such as dextran, produced by Leuconostoc and Lactobacillus spp., have a severe impact on factory throughput and sugar quality. This study aimed to determine the phylogenetic relationships between gum‐producing Leuconostoc and Lactobacillus bacteria which were isolated from various locations in a sugarcane processing factory at times when low‐ and high‐dextran raw sugar, respectively, were produced. Phylogenetic analysis of 16S rRNA gene sequences grouped 81 isolates with the type strains of Leuconostoc mesenteroides (subspp. mesenteroides, dextranicum, and cremoris), Leuconostoc pseudomesenteroides, Leuconostoc lactis, and Leuconostoc citreum, respectively. Forty‐three isolates clustered with the type strain of Lactobacillus fermentum. The phylogenetic relatedness of the isolates was determined by sequencing and analysis of the housekeeping genes rpoA and dnaA for Leuconostoc spp. and the pheS and tuf genes for the Lactobacillus spp. The rpoA gene proved discriminatory for the phylogenetic resolution of all of the isolated Leuconostoc spp. and the dnaA housekeeping gene was shown to be effective for isolates clustering with the type strains of Leuc. mesenteroides and Leuc. citreum. None of the loci examined permitted differentiation at the subspecies level of Leuc. mesenteroides. Single‐locus analysis, as well as the concatenation of the pheS and tuf housekeeping gene sequences, yielded identical phylogenies for the Lactobacillus isolates corresponding to L. fermentum.

The authors differentiated Leuconostoc from Lactobacillus based on the assumption that lactobacilli do not produce dextran from sucrose, despite reports to the contrary (Duncan & Seeley, 1963, 1965Pederson & Albury, 1955). Previous attempts by sugar technologists to identify spoilage bacteria in sugarcane processing factories have all been hampered by the lack of available identification methods of high discriminatory power.
Dextran-producing strains of Leuconostoc mesenteroides have been implicated in the reduction of factory throughput and quality of the produced sugar (Eggleston, Morel du Boil, & Walford, 2008). Some strains of Leuc. mesenteroides produce as much as onepart dextran from 40 parts of sucrose after only 6 hr (Cerutti de Guglielmone, Diez, Cárdenas, & Oliver, 2000). Apart from an increase in viscosity, other metabolites produced during the degradation of sucrose reduce the purity of the cane juice and thus also sucrose recovery. Impure juice requires longer boiling times, leading to higher sucrose inversion losses. The impurities reduce evaporation rates, and sugar crystals take longer to form (Godshall, Legendre, Clarke, Miranda, & Blanco, 1996;Jimenez, 2005). Dextran shows high (20%) transfer from juice to crystal, resulting in high carryover from the factory to the refinery. Refiners and buyers of raw sugar prefer the product to have low levels of dextran (<100-150 mg/kg), even if the purchase contract does not specify dextran content (Ravnö & Purchase, 2005). Economic losses due to microbial activities are, therefore, not limited to the direct loss of recoverable sucrose and indirect loss due to reduced factory throughput, but also finding financially attractive markets for high-dextran raw sugar (Moodley & Khomo, 2018).
A recent study by Nel, Davis, Endo, and Dicks (2019a) used phylogenetic analysis of 16S rRNA gene sequences for the profiling of gum-producing bacteria which were isolated from various locations in a South African sugarcane processing factory at times when lowand high-dextran raw sugar, respectively, were produced. Although 16S rRNA gene sequences have been widely used as a phylogenetic marker in bacterial taxonomy, the method has limitations and is not reliable in distinguishing species and subspecies with high sequence similarities (Jeon et al., 2017). Leuc. mesenteroides and Leuconostoc pseudomesenteroides are good examples. The two species share almost identical 16S rRNA gene sequences, with differences in only 5 of the 1,483 nucleotides (Martinez-Murcia & Collins, 1990).
Comparisons among housekeeping gene sequences are commonly used to overcome the limitations of 16S rRNA coding region sequencing (Chelo, Ze-Ze, & Tenreiro, 2007;De Bruyne et al., 2007;Naser et al., 2007;Yu et al., 2012). Single protein-coding genes do not reflect general phylogenetic relationships due to potential horizontal gene transfer (HGT) or lateral gene transfer (Gogarten, Doolittle, & Lawrence, 2002;Macheras et al., 2011). Multiple genebased phylogenies were introduced which have been used more frequently to overcome the bias caused by single gene sequence-based phylogenies (Glaeser & Kämpfer, 2015). Concatenation of several housekeeping genes may reduce the weight of HGT, and it could accurately locate taxonomic positions for closely related species and strains (Glaeser & Kämpfer, 2015). Ricciardi, Storti, Zotta, Felis, and Parente (2020) Ricciardi et al. (2020) suggested that the genome-based classification of Leuc. mesenteroides should be supported by comparative metabolic diversity studies to identify molecular markers, such as taxonomically and functionally relevant genes, for rapid detection and discrimination of strains.
Previously, we reported on the phylogenetic identification of 430 gum-producing isolates from harvested sugarcane and sampled from a South African sugarcane processing factory, based on partial 16S rRNA gene sequence analysis (Nel et al., 2019a). A large number of these isolates (47%) were identified as Weissella cibaria and Weissella confusa, respectively (Nel, Davis, Endo, & Dicks, 2019b).
Most of these bacteria were isolated from the prepared (shredded) sugarcane at a time when high-dextran raw sugar was produced. The only bacteria isolated from the juice screen and the mixed juice tank (24% of the isolates) were identified as Bacillus amyloliquefaciens and Bacillus subtilis, respectively (Nel, Davis, Endo, & Dicks, 2019c). Again, most of these bacilli were isolated when high concentrations of dextran were reported in the raw sugar. Only 19% of the isolates were identified as Leuconostoc spp. and 10% as Lactobacillus spp. based on initial 16S rRNA gene sequencing analysis (Nel et al., 2019a). This study aimed to determine the phylogenetic relationships between the isolated Leuconostoc and Lactobacillus bacteria based on the phylogenetic relatedness of selected housekeeping gene sequences.

| Isolation of gum-producing bacteria
Samples of shredded (prepared) sugarcane and samples from the diffuser sump, juice screen (Dutch State Mines; DSM screen), mixed juice tank (MJ tank), filtrate, mud trough, and syrup tank in a South African sugarcane processing factory were collected and screened for the presence of gum-(polysaccharide) producing bacteria (Nel et al., 2019a). Once-off samples at each sampling location were taken during September 2013, when low-dextran concentrations (<70 ppm) in the produced raw sugar were reported. This was repeated in November 2013, when high-dextran concentrations (>500 ppm) in raw sugar were found. Cane samples (10 g each) were added to 100 ml phosphate-buffered saline (PBS; Green & Sambrook, 2012) and incubated on a rotary shaker (30°C, 150 rpm) for 1 hr. Liquid samples collected from each of the sampling points and PBS-cane suspensions were serially diluted in PBS. Serial dilutions were streaked onto modified dextransucrase-inducing agar with the following composition: sucrose 100 g/L, peptone 20 g/L, KH 2 PO 4 20 g/L, agar 15 g/L, and R-salts (4% MgSO 4 ·7H 2 O, 4% NaCl, 0.2% FeSO 4 ·7H 2 O, and 0.2% MnSO 4 ·H 2 O) 5 ml (Tsuchiya et al., 1952). Plates were incubated at 30°C for 14-18 hr. Visual screening of colonies with a glistening and slimy appearance on the dextransucrase-inducing medium was carried out to select for gum-producing bacteria. A total of 430 colonies were isolated and streaked to purity on modified dextransucrase-inducing agar. From these plates, a single colony was inoculated into 5 ml MRS broth (Biolab, Merck South Africa), and the cultures were incubated on a shaking incubator (150 rpm) for 14-18 hr at 30°C. Cells were harvested by centrifugation (16,000 g, 25°C, 2 min), resuspended in sterile glycerol (200 μl; 50%, v/v), and stored at −70°C.

| Genomic DNA extraction
An aliquot of ten microliters from stock cultures of previously isolated (Nel et al., 2019a) Leuconostoc (81 isolates) and Lactobacillus (43 isolates) was inoculated into 5 ml sterile MRS broth (Biolab, Merck) and incubated for 16 hr at 30°C on a rotary shaker (150 rpm). Cells were harvested (16,000 g, 25°C, 2 min) and genomic DNA extracted using the GeneJET Genomic DNA Purification kit (Thermo Scientific, Inqaba Biotechnical Industries) according to the manufacturer's instructions. Purified DNA was suspended in 50 μl elution buffer and used as a template in amplification reactions.

| Amplification of the 16S rRNA genes, and housekeeping genes rpoA, dnaA, pheS, and tuf
Genomic DNA was used as a template to amplify sequences of the 16S rRNA genes of all species, rpoA and dnaA genes of Leuconostoc spp., and pheS and tuf genes of Lactobacillus spp. using the primers listed in Table 1. Reactions were carried out in 50 μl, containing 10 pmol of each primer, 200 µM of each deoxynucleoside triphosphate (Thermo Scientific), 10 µl of 5× One Taq Standard Reaction buffer, 1.25 U One Taq Hot Start DNA polymerase (Thermo Scientific), and 100 ng template genomic DNA. PCR reactions were performed in a programmable thermal cycler (MultiGene OptiMax, Labnet International, Whitehead Scientific) with an initial denaturation step (94°C, 30 s), followed by 30 cycles of denaturation (94°C, 30 s), primer annealing, and elongation (see Table 1). Cycling was completed by a final elongation step (68°C, 10 min), followed by cooling to 4°C. The amplified fragments were purified using the DNA Clean and Concentrator™-25 kit (Zymo Research, Inqaba Biotechnical Industries) according to the manufacturer's instructions.
Sequence similarity searches were performed using the Basic Local Alignment Search Tool (BLAST) algorithm (Altschul, Gish, Miller, Myers, & Lipman, 1990 Sequences were aligned with ClustalW (Thompson, Higgins, & Gibson, 1994), as implemented in the BioEdit Sequence Alignment Editor program (Hall, 1999). A data matrix for each alignment was created for the representative sequences of strains at each sampling location and sampling time. Phylogenetic analyses were conducted using the Molecular Evolutionary Genetics Analysis (MEGA) version 7.0 software (Kumar, Stecher, & Tamura, 2016).
Evolutionary histories were inferred using the maximum likelihood method with the Kimura 2-parameter model (Kimura, 1980) for 16S rRNA sequence analyses, Lactobacillus spp. tuf sequence analyses and Lactobacillus spp. pheS-tuf concatenated sequence analyses.
The strengths of the internal branches of the resultant trees were statistically evaluated by bootstrap analysis (Felsenstein, 1985) with 100 bootstrap replications.
One-hundred and twenty-four isolates of Leuconostoc and Lactobacillus spp. grouped into five distinct clusters based on 16S rRNA gene sequence analyses (Figure 1). Of the 124 isolates, 81 were classified as members of the genus Leuconostoc.
TA B L E 2 GenBank accession numbers of the sequences as determined in this study for representative Leuconostoc and Lactobacillus strains for each sampling location F I G U R E 1 Phylogenetic tree based on partial 16S rRNA gene sequences of Leuconostoc and Lactobacillus species isolated from five sampling points in a South African sugarcane processing factory. Isolates from the sampling time when low-dextran content was observed in raw sugar are labeled with a circle (•) and those when high dextran in raw sugar was reported with a square (■). The tree was constructed using the maximum likelihood method with MEGA 7.0 software, and representative isolates from each sampling point are shown, with the number of isolates indicated in brackets. Sequence data of reference strains were from GenBank. Genetic distances were computed by Kimura's 2-parameter model (Kimura, 1980  None of the loci examined (Figures 1 and 2a,b), nor the concatenated rpoA-dnaA sequences (Figure 2c), allowed discrimination between subspecies within Leuc. mesenteroides.

| Phylogenetic analyses of amplified pheS, tuf, and pheS-tuf concatenated gene sequences of Lactobacillus spp.
The 43 lactobacilli in this study clustered with L. fermentum in the phylogenetic trees inferred from partial pheS (Figure 3a), tuf F I G U R E 2 Phylogenetic trees based on partial rpoA (a), dnaA (b), and rpoA-dnaA concatenated (c) gene sequences of Leuconostoc species isolated from five sampling points in a South African sugarcane processing factory. Isolates from the sampling time when low-dextran content was observed in raw sugar are labeled with a circle (•) and those when high dextran in raw sugar was reported with a square (■). The tree was constructed using the maximum likelihood method with MEGA 7.0 software, and representative isolates from each sampling point are shown, with the number of isolates indicated in brackets. Sequence data of reference strains were from GenBank. Bootstrap values over 50% (based on 100 replications) are shown at each node. Bar, % estimated substitution per nucleotide position

| D ISCUSS I ON
The rpoA housekeeping gene has previously been successfully applied for the phylogenetic resolution of Leuconostoc spp. (De Bruyne et al., 2007;Rahkila, De Bruyne, Johansson, Vandamme, & Björkroth, 2014). Although strains of Leuconostoc spp. could be differentiated to species level by phylogenetic analyses of their rpoA housekeeping genes, the phylogenetic resolution was not higher compared to 16S rRNA gene analyses. Comparison of dnaA gene sequences differentiated isolates that clustered with the type strains of Leuc. mesenteroides and Leuc. citreum. None of the genes analyzed differentiated between subspecies of Leuc. mesenteroides.
Reasons for this have been ascribed to an inadequate number of available genomes and genomic diversity studies and fewer strains of Leuc. mesenteroides subspp. cremoris, dextranicum, and jonggajibkimchii studied compared Leuc. mesenteroides subsp. mesenteroides (Ricciardi et al., 2020). Only a few strains of Leuc. mesenteroides subspp. cremoris and dextranicum are listed in microbial culture collections (e.g., ATCC, DSMZ, LMG), and for some of them, the isolation source is unknown, which limits information on strain diversity. The description of Leuc. mesenteroides subsp. jonggajibkimchii was based on phenotypic and genomic features of a single isolate (DRC1506 T , Jeon et al., 2017). Ricciardi et al. (2020) suggested that more strains need to be studied to reach a confident separation at the subspecies level (Ricciardi et al., 2020). For this reason, Leuc. mesenteroides subsp. jonggajibkimchii was not included in any of the phylogenetic analyses presented in this work. Ricciardi et al. (2020) used multiplex PCR to differentiate dextran-producing strains of Leuc. mesenteroides subsp. mesenteroides from nondextran producing strains of Leuc. mesenteroides subsp. cremoris. The method was based on the detection of L-arabinose isomerase, dextransucrase, and PTSsorbose-transporter subunit IIC genes. Although multiplex PCR could not resolve ambiguous identification of Leuc. mesenteroides subspp. dextranicum and jonggajibkimchii strains, the method may be of value to identify gum-producing Leuconostoc spp. isolated from sugarcane processing factories.
The phylogenetic analyses of pheS (encoding phenylalanyl-tRNA synthase) and tuf (encoding elongation factor Tu) genes have proven to be a valuable tool for the taxonomic resolution of Lactobacillus spp. (Chavagnat, Haueter, Jimeno, & Casey, 2002;Naser et al., 2007;Sarmiento-Rubiano et al., 2010;Ventura, Canchaya, Meylan, Klaenhammer, & Zink, 2003;Yu et al., 2012). In this study, single-locus analysis, as well as a concatenation of the pheS and tuf housekeeping gene sequences, yielded identical phylogenies for the Lactobacillus isolates corresponding to L. fermentum. This was in agreement with the 16S rRNA gene sequence analysis.

F I G U R E 3
Phylogenetic trees based on partial pheS (a), tuf (b), and pheS-tuf concatenated (c) gene sequences of Lactobacillus species isolated from three sampling points in a South African sugarcane processing factory. Isolates from the sampling time when low-dextran content was observed in raw sugar are labeled with a circle (•) and those when high dextran in raw sugar was reported with a square (■). The tree was constructed using the maximum likelihood method with MEGA 7.0 software, and representative isolates from each sampling point are shown, with the number of isolates indicated in brackets. Bootstrap values over 50% (based on 100 replications) are shown at each node. Bar, % estimated substitution per nucleotide position The quality of sugarcane reaching the factory has to be tested regularly. This is a major challenge since there is no rapid, reliable, and inexpensive method available to detect the level of cane deterioration (Eggleston & Harper, 2006). The modified haze method (Anon, 2015), used by the sugarcane processing factory samples were taken from, determines the overall gum content in raw sugar and not dextran. In this study, we have analyzed the sequences of housekeeping genes to identify dextran-producing strains of Leuconostoc and Lactobacillus spp. from various locations in the sugarcane processing factory. The presence of dextran-producing strains served as an indicator of dextran production and cane deterioration. Leuconostoc spp. and L. fermentum produce dextran and other metabolic products, including mannitol, lactic and acetic acids, and ethanol (Daeschel, Andersson, & Fleming, 1987;Eggleston, Legendre, & Tew, 2004), and are thus considered spoilage organisms. Although dextran is considered to be the most detrimental product to the factory because it is a high-viscosity polymer, Leuconostoc and Lactobacillus spp. are also capable of producing other polymers such as levan and alternan (Dutta, Das, & Goyal, 2012;Kralj et al., 2004;Naessens, Cerdobbel, Soetaert, & Vandamme, 2005). The formation of these polymers may be underestimated as contributors to impeding high-viscosity problems in sugarcane processing, mainly because of the nonspecific nature of the dextran quantification method (Anon, 2015) used in the sugar industry. These bacterial metabolites may have a severe impact on the quality and quantity of produced sugar.
The number of Leuconostoc and Lactobacillus bacteria which were isolated when low-dextran raw sugar was produced (40 and 23 isolates, respectively) was similar to the numbers of these bacteria when high-dextran raw sugar was produced (41 and 20 isolates, respectively). Contrary to previous reports (Egan, 1966;Solomon, 2000), Leuconostoc spp. were not the major gum-producing bacteria isolated from sugarcane (Nel et al., 2019a). This study showed that the isolate closely related to Leuc. citreum was the only Leuconostoc bacterium isolated from shredded (prepared) sugarcane, and it was shown previously (Nel et al., 2019b) that W. confusa and W. cibaria were the most prevalent gum-producing bacteria on the prepared cane. Similar numbers of isolates clustering closely with L. fermentum were isolated from the diffuser sump, filtrate, and mud trough at both sampling times, respectively. Isolates that were phylogenetically related Leuc. lactis were the most prevalent in the filtrate at a time when high-dextran raw sugar was produced. The majority of bacteria isolated from the mud belonged to species clustering with Leuc. mesenteroides (low-dextran raw sugar) and Leuc. lactis (high-dextran raw sugar), and isolates related to Leuc. mesenteroides were the dominant bacteria isolated from the syrup tank at both sampling times.
Correct process control, especially of high-temperature streams, is critical to prevent microbial growth in a sugarcane processing factory. In this study, filtrate temperatures of 58°C and 29°C were recorded when sampled at times of low and high-dextran concentrations in raw sugar, respectively. Filtrate temperatures are usually around 60°C. Factory staff acknowledged that the low filtrate temperature recorded during the second sampling was due to a processing error. At this time, strains of Leuc. lactis and L. fermentum were isolated. Lactobacillus fermentum dominated the filtrate sample taken at the first sampling.
The temperature of the mud in the mud trough at this time (35°C) was much lower compared to the second sampling (64°C), possibly due to stoppages and longer retention times of the mud in the trough, resulting in cooling of the mud. A considerable number of Leuc. mesenteroides strains (31% of the total number of strains isolated during the first sampling) were from mud at 35°C.
On the contrary, Leuc. lactis was the major gum-producer in the mud during the second sampling when the temperature was higher (64°C). Leuc. lactis has a higher heat resistance than Leuc. mesenteroides (Logan & De Vos, 2009). Although the filtrate is recirculated to the mixed juice tank, none of the gum-producing bacteria isolated in the filtrate were detected in the juice sampled from the mixed juice tank. This is presumably due to the high temperatures (67°C and 73°C, respectively) recorded for juice samples, which allowed the growth of endospore-forming Bacillus species, but not Leuconostoc and Lactobacillus spp. (Berendsen et al., 2016;Logan & De Vos, 2009;Warth, 1978).

| CON CLUS IONS
Dextran is an unwanted bacterial metabolite in sugarcane processing, leading to reduced factory throughput, quality, and quantity of the produced sugar. This study showed that the number of lactobacilli at the various locations at both sampling times was similar and all species were found to be related to L. fermentum. However, the diversity of the leuconostocs was found to vary depending on the temperature of the location from which they were isolated. Correct process control of high-temperature factory streams is therefore critical to limit microbial growth and gum-formation in a sugarcane processing factory.
The phylogenetic relationships, based on housekeeping gene sequence analyses, of Leuconostoc and Lactobacillus species isolated from various unit operations of a sugarcane processing factory were established at a time when low-and high-dextran raw sugar, respectively, were produced. Analyses of rpoA sequences proved as effective as 16S rRNA gene sequence analyses to determine the phylogenetic relationships between Leuconostoc spp. isolated from sugarcane. Comparison of dnaA sequences differentiated isolates that clustered with the type strains of Leuc. mesenteroides and Leuc. citreum. Clear differences were recorded between Lactobacillus spp. and the type strains of Lactobacillus spp. when sequences of pheS and tuf were compared. Although the housekeeping genes did not prove more discriminating compared to 16S rRNA gene sequence analyses, this study illustrated the potential of gene-based methods as an alternative to phenotypic methods to differentiate lactic acid bacteria in the sugarcane industry.

ACK N OWLED G M ENTS
The authors thank Dr Deborah Sweby (South African Sugarcane Research Institute, Mt Edgecombe, South Africa) for DNA sequencing analyses. No funding declared.

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

E TH I C S S TATEM ENT
None required.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data are provided in full in the results section of this paper. The relevant DNA sequences were deposited in GenBank, and accession numbers are listed in Table 2.